Automated and integrated process using artificial intelligence and artificial intelligence-based teams for data analysis, early detection and definition of preventive measures against crime

Abstract

The present invention relates to a universal process based on full respect for human rights, ethics and responsibility. The present invention pertains to the legal and technological fields and may be applied to the Brazilian Government, foreign governments, national and international universities, public and private enterprises, as well as national and international research centres. Said invention comprises processes referred to as IAPs, which integrate artificial intelligence-based teams and which participate in decision-making. By means of artificial intelligence, said processes are capable of creating new preventive methods, improving existing preventive methods, and performing continuous monitoring and control so as to prevent violations of human rights. The present invention is intended to reduce crime in Brazil and worldwide and is further capable of analysing war crimes in order to identify preventive measures for the establishment of peace agreements.

Description

Automated and integrated process using artificial intelligence and teams composed of artificial intelligence for data analysis, early detection, and definition of preventive measures against crimes. Field of invention

[0001] The present invention belongs to the field of processes for enhancing and creating methods for crime prevention. The invention process relates to the legal and technological areas. It has applications for the Brazilian Government, Governments of Foreign Countries, National Universities, International Universities, Public Companies, Private Companies, National Research Centers and International Research Centers.

[0002] Fundamentals of the invention

[0003] This process aims, through the use of teams formed by artificial intelligence, to enhance preventive methods related to crime by Criminal Line and to enhance / create preventive methods to avoid crimes from happening in the future. This patent possesses techniques that will be able to benefit society by preventing crime.

[0004] We will act primarily in a preventive manner. By understanding how violence occurs and its root cause, combined with the application of Artificial Intelligence, we can act "ex ante," uniting the past, present, and future. From the past, we will use historical data; from the present, we will analyze currently applied preventive initiatives and methods; for the future, we will predict patterns of behavior and events, and identify which preventive methods should be enhanced or created to reduce crime. This robust process integrates Law and Technology, connecting the Criminal field and Artificial Intelligence with a focus on Human Dignity. Its main pillar is respect for Human Rights and the main International Treaties and Decrees. To comply with the Law, intrinsically connected values ​​such as fundamental principles and human rights have been incorporated into the process.

[0005] It is important to emphasize that the present invention does not aim to identify crimes in order to create preventive methods for predictive policing.

[0006] To understand the basis of the invention, it is important to understand the triad: Crime, Technological Evolution, and Artificial Intelligence.

[0007] Crime is part of human history. Great philosophers like Paul-Michel Foucault conducted political studies on criminology, addressing how social structures are formed within the historical context and thought of each era. Throughout the development of society, crime has been analyzed and punished in different ways according to the context of each era. The Italian Cesare Beccaria Bonesana was the first to assert that torture, one of the punitive methods of past centuries, should be eliminated. The Second World War was a major landmark of atrocities committed by humankind, where discrimination was manifested in all its forms. The genocide committed before the world drew the attention of major nations, thus leading to the creation of the Universal Declaration of Human Rights by the United Nations (UN).This milestone offered a new perspective on humanity through the creation of laws guaranteeing human rights, such as the 1988 Federal Constitution. Crime in Brazil and worldwide has been increasing drastically. Data from sources such as the Brazilian Institute of Public Security (ISP) and the Brazilian Yearbook of Public Security confirm this unfortunate reality. Robbery resulting in death, homicide, bodily injury followed by death, theft, swindling, racial slurs, femicide, stalking, and cybercrimes are some examples of crimes that are increasing daily. Countries meet in confederations and hold international meetings, but even so, they fail to identify effective preventative methods capable of helping humanity.

[0008] Just as the 1988 Constitution, among other laws, emerged, so too has technological advancement developed throughout history. Looking at the historical context of computers, we've seen the creation of mechanical devices dating from 500 BC to 1880, followed by the development of electromechanical devices, vacuum tube computers, transistorized computers, integrated circuit computers, mainframe computers, and personal computers. Just as their physical components were developed, their logical components also underwent significant evolution. Moving from one technology to another was a common challenge faced by technology professionals, from migrating from programming languages ​​like COBOL, FORTRAN, CLIPPER, PASCAL, among others, to object-based languages, and even developing inference tools for Artificial Intelligence that evolved from rule-based systems to frame-based and object-based systems.

[0009] These challenges, which are overcome daily by technology professionals, have also become a challenge for legal professionals since technological advancements entered the legal world, with the digitization of processes transforming them into digital processes. Legal software has also been developed to streamline daily work. Therefore, these professionals have needed to learn how to use systems to keep up with legal updates, consult case law, review precedents, consult decisions, generate legal documents, and perform other tasks.

[0010] Artificial Intelligence (AI) has also evolved from selection by inference, where deductions were made through chaining, to Generative Artificial Intelligence. For decades, the goal of developers and researchers has been to create a process capable of combining the best of the human brain and the best of machines, that is, the best synergy between the two, thus evolving Artificial Intelligence. Considered the Fourth Industrial Revolution, it is being applied in various areas such as: Agribusiness, Environment, Law, Capital Markets, Health, Security, Human Resources, Entertainment, Education, Transportation, among others. In this way, this technology is affecting the world in various aspects, bringing benefits to society.

[0011] To better understand what AI is, we need to address the definition of data. Data is the smallest unit of information. This unit, when analyzed alone, does not generate information or context. When analyzed together with other data, it generates interpretive information and contexts, identifying behavioral patterns for decision-making or process automation. It is capable of simulating thought, storing far more information than the human brain and processing information at high speed. Data is already considered the greatest of all riches, because through it we can generate contextualized and strategic information for various specialties. For this reason, Big Tech leaders, that is, leaders who dominate the technology market, fiercely compete to obtain user data, extracting it from a wide variety of sources such as applications, social networks, and websites.A society already experiencing the Internet of Things and Smart Cities, where daily life involves connecting to the internet in various ways, provides data through clicks and access. This data is analyzed by algorithms, generating strategic information such as: shopping preferences and website access, opinions on diverse areas like politics, medical exams, and other information. In this way, companies profit increasingly from these strategies. Furthermore, besides being used for profit, this powerful technology is being used to develop war strategies. AI has several divisions, and we can highlight some, such as: Recommendation Systems, Robotics, Computer Vision, Natural Language Processing, Machine Learning, and Generative AI.In recommendation systems, it is used to identify behavioral patterns, thus suggesting new options according to user preferences. In robotics, it is applied, for example, in the development of autonomous cars and military robots. In computer vision, it is used to assist in the preparation of medical reports, facial recognition, detection of firearms, and drones using satellite imagery. In natural language processing, it is used for sentiment analysis, translation, interpretation, and summarization of texts. With machine learning, we can predict diseases, environmental disasters, and analyze capital markets. Generative AI can generate any type of content.

[0012] While major leaders seek profitability through user preference or pursue war strategies, the IAPs (Preventive Artificial Intelligence) process offers a new perspective on this powerful technology: "the pursuit of crime prevention and, consequently, greater human dignity for our society." We need to illuminate the path so that society can live with dignity. We are living through a moment of "Technological Enlightenment" where major technological transformations are occurring and will continue to occur, impacting society.

[0013] Based on fundamental principles, the IAPs (Innovation and Planning System) was developed with full respect for Human Rights and conducted with Ethics and Responsibility. There is undoubtedly a great concern regarding discriminatory biases that can be generated by data or algorithmic processing. For this reason, the IAPs has processes in place to prevent the perpetuation of these biases, maintaining closed borders and permanent monitoring and control to ensure that no violations of laws and human rights occur. The IAPs will follow protocols of best practices and requirements that may be requested by competent authorities in the future, respecting commercial and industrial secrets. Technological solutions originating from the IAPs process will follow best practices required by law.

[0014] Context for some words that will be part of this document: AI: Artificial Intelligence IAP: Preventive Artificial Intelligence against crime. Technological Enlightenment: Technological advancement that affects society. Frontier: A frontier formed by algorithms. Prediction: A process that analyzes data and models to predict future events. AI Inference: Mechanisms and techniques encompassing processes performed by Artificial Intelligence that are not limited to: Data pre-processing, model creation, integration, training and testing, algorithms. The AI ​​inference techniques applied in the lAPs process may be altered as technology evolves; that is, the techniques mentioned in the present invention are not limited to those currently known. Compliance: Conformity with the Laws, Ethical Values, and Values ​​defined for the analysis of the crime in question. Direct Discrimination: This occurs when the algorithm uses discriminatory features to make decisions. Indirect discrimination: discrimination that occurs when an apparently neutral criterion can lead to disadvantage for people belonging to a specific group, or place them at a disadvantage. IAP Criminal Line: IAP Red Line: Crimes committed against the Person. IAP Black Line: Crimes committed against Sexual Dignity. IAP Marine Line: Crimes committed against property. IAP-Orange Line: Crimes committed against Intellectual Property. IAP-Gray Line: Crimes committed against Religious Sentiment and Against Respect for the Dead. IAP-Yellow Line: Crimes committed against the Family. IAP Green Line: Crimes committed against the Environment. IAP Blue Line: Crimes against public administration, Crimes against public peace, Crimes against public faith, and Crimes against public safety. IAP-Brown Line: Other Crimes IAP White Line: One or more crimes related to the following lines: Red, Black, Navy, Orange, Gray, Yellow, Green, Blue, and Brown that were committed during wars. The Ethics Committee will have a special focus on identifying new preventive methods capable of facilitating peace agreements.

[0015] The Crime Trail correlated to Crimes is capable of identifying connections between crimes and thus providing greater efficiency in enhancing and developing preventive methods.

[0016] There are some patents abroad that involve crime prediction: IN202431093827, IN202241006900, IN202441087476, WO2019103205, CN114118084, IN202441002553, IN202341086172, CN114742282, CN118820708, IN202441032381, IN202311047874, CN105512995, IN202411057965, IN202421058767, IN202341064022, IN202341035525, CN108805345, IN202431093827, KR101830522B1, US11436510B1 and US11087614B2.

[0017] South Korea's patent KR101830522B1 performs crime prediction by region, integrating multiple domains such as population, economic, educational, housing, climate, and image domains; and aims to predict crimes using deep learning techniques.

[0018] US patent 11436510B1 developed a crime prediction system for predictive policing. In contrast, the present claim does not aim to create predictive systems to increase police monitoring, but rather to enhance preventive methods (unrelated to increased policing) and / or create preventive methods not yet identified for crime prevention.

[0019] US patent 11087614B2 from the United States developed a personalized and adaptive machine learning method for crime alerts that presents a set of customized corrective and preventive actions for students, based on at least one student attribute data point. A set of student safety performance statistics is used to generate an optimized safety management strategy for the school security team, with the goal of reducing school crime. The other patents IN202431093827, IN202241006900, IN202441087476, WO2019103205, CN114118084, IN202441002553, IN202341086172, CN114742282, CN118820708, IN202441032381, IN202311047874, CN105512995, IN202411057965, IN202421058767, IN202341064022, IN202341035525, CN108805345 were also created for predicting the increase in crimes for the State's Public Security with The objective of increasing predictive policing.

[0020] Patent IN202431093827 developed a decision support system to increase the efficiency of investigations and resource allocation for increased patrols during crime-related hours. We can observe in the patent drawings that the analysis is performed by crime. There is no correlation with criminal areas, preventative methods, or the operation of teams formed by Artificial Intelligence.

[0021] There is no Brazilian patent for crime prevention that does not aim at predictive policing. Among the patents related to Artificial Intelligence listed below, none have processes similar to or described as the present application: BR1020240043340, BR1020240043332, BR1020240029143, BR1020230163645, BR1020230145116, BR2020230118360, BR1020230107508, BR1020230098142, BR1120240230503, BR1020230068871, BR1020230053912. BR1120240173860, BR1120240181561, BR1020230034500, BR1120240169014 ;BR1120240163555, BR1020230011780, BR1120240154440, BR1020230003630 ; BR1020220264481, BR1020220253218, BR1120240105393, BR1020220234221; BR1120240069656, BR1020220202664, BR1120240062554, BR1020220187657; BR1020220184534, BR1120240040771, BR1120230277523, BR1020220148945; BR1020220147981, BR2020220139135, BR1020220136327, BR1020220125368; BR1020220125082, BR1020220116415, BR1020220108706, BR1120230249023; BR2020220098285, BR1120230229871, BR1120230201470, BR1120230201985; BR1120230186315, BR1120240153584, BR1120230144817, BR1120230139180; BR1120230132495, BR1120240123596, BR1120230172772, BR1020210235411; BR1120230100070, BR1020210232102, BR2020210230733, BR1120240081664 BR1120230065445, BR1020210199768, BR1020210189428, BR1020210168943.BR1120220266421, BR1020210105496, BR1020210100907, BR1020210042923 ; BR1020210032553, BR1120220164157, BR1120220164955, BR1120220164165 ;BR1120220164092, BR1120220137095, BR1120220136609, BR1120220198612: BR1120220129882, BR1120230126266, BR1120220113862, BR1120230055342 BR1020200223399, BR1120220070918, BR1020200170767, BR1120220174934: BR1020200162470, BR1020200148869, BR1020200133942, BR1020200115944: BR1020200087550, BR1120210187014, BR1120210186956, BR1020200077945: BR1120200264555, BR1120200264083, BR1120200264261, BR1120200264334: BR1120210181040, BR1020200047418, BR1120210102981, BR2020190203450: BR1120200253448, BR1120210104682, BR1120200180130, BR1020190049669: BR1020190043849, BR1020180747916, BR2020180745560, BR1120190271872: BR1020180023675, BR1020170253236, BR1020170169103, BR1120190114718 ; BR1120180150142, BR1020160243815, BR1120180061948, BR2020160200216: BR1020160056381, BR1020150191308, BR1120150328881, BR1020120281112: BR1020120281120, PH 104441-1, PI1004529-5, PI1015255-5, PI0604102-7 PI0417217-5, PI0303623-5, PI0205900-2, PI0208257-8, PI0208256-0, PI0208307-8, PI0207087-1, PI9911605-7.

[0022] In risk management, patent US11087614B2 refers to a task risk management system. In this patent US11087614B2, the task risk management system comprises: a user terminal on which a private program is installed to provide a task risk management service over a network; and a server that provides the task risk management service to the user, with information presented in a graph and a diagram, through the private program installed on the user terminal.

[0023] It is important to emphasize that most scientific articles in the field deal with the application of Artificial Intelligence in crime prediction and predictive policing, or are limited to presenting criticisms of Artificial Intelligence systems without proposing concrete solutions — unlike the patent presented here.

[0024] These patents only identify increases in crime so that Public Security can increase predictive policing, without identifying or enhancing preventive methods, creating preventive methods, or having teams formed by artificial intelligence acting autonomously. In contrast, the present invention does not aim to create predictive systems for police monitoring, but rather to enhance preventive methods (not related to increased policing) and / or create preventive methods that do not yet exist. It is important to emphasize that patents that make predictions about crime increases for predictive policing, considering region, race, and other data, naturally lead to discrimination in state action. This is precisely what the present patent does not propose.

[0025] Although the world's attention is focused on Artificial Intelligence, no one has been able to develop a crime "prevention" process that considers human dignity. The IAPs process has universal principles that can be applied in any country.

[0026] Innovation:

[0027] Enhancing preventive methods through Artificial Intelligence and AI-powered teams to reduce crime, thus achieving greater effectiveness and efficiency: Through AI-powered teams, we will enhance preventive methods against crime.

[0028] Creating non-existent preventive methods through Artificial Intelligence and through AI-powered teams to reduce crime, thus achieving greater effectiveness and efficiency: Through AI-powered teams, we will create preventive methods that do not yet exist.

[0029] Creation of AI-powered teams that work in an integrated manner, achieving greater efficiency in all phases of the processes: Through AI, an Ethics Team is created for each Criminal Line, capable of making decisions to prevent violations of Ethics and Human Rights. Through AI, an AI Expert is created, responsible for preparing Legal Technical Opinions on actions that may or may not succeed, protecting the Democratic Rule of Law and Human Rights, achieving greater precision and efficiency in decision-making regarding the application of prevention. Through AI, a risk team is created, capable of making decisions by analyzing risks.

[0030] Analysis of the Legal Framework for the use of preventive methods generated by teams formed by Artificial Intelligence: With the use of teams formed by Artificial Intelligence and the integration of data from the Legal Framework, the legality of the preventive methods will be guaranteed.

[0031] Risk analysis performed by a team powered by Artificial Intelligence: Through the operation of integrated teams powered by Artificial Intelligence, risk analyses are performed, thereby achieving greater efficiency, effectiveness, and efficacy in the results.

[0032] Continuous Monitoring and Control by AI-Powered Teams: AI-powered teams provide continuous monitoring, ensuring greater agility in all areas of operation, legal compliance, and increased efficiency in prevention and results. By applying preventative methods, we enable improvement, societal well-being, and greater human dignity.

[0033] Removing discriminatory biases through AI-powered teams: By using AI-powered teams and applying machine learning techniques to address discriminatory biases, the process removes biases, enabling non-discriminatory actions.

[0034] Report generation by AI-powered teams: AI-powered teams generate reports with greater efficiency, providing the transparency that is a key requirement for AI processes.

[0035] Technical Details of the Invention with the involvement of teams formed by Artificial Intelligence:

[0036] Data Processing and Exploration: The goal of this stage is to perform both the preparation and initial exploration of the data to detect data patterns. Processing is carried out so that the data can be used in the models. Data collection, cleaning, normalization, coding / conversion, tokenization, repository indexing, stemming / lemmatization applications, and tokenizer configuration for data inference and exploration are performed.

[0037] Preparation of AI Models for the Process: The goal is to consolidate different types of AI and enable multimodal analyses. This stage involves training and testing. It also includes the integration of multimodal models that combine Natural Language Processing, Machine Learning, and Generative Artificial Intelligence, information retrieval for searching, and information integration.

[0038] Model Application Process: The objective of this step is to apply the models and verify if they can correctly capture predictive patterns. This step involves temporal analysis to check for seasonality and trends. Overfitting and underfitting analysis is also performed to identify how to enhance learning, information retrieval for searching, and information integration. For example, algorithms such as ARIMA and Temporal Transformers are applied.

[0039] Enhanced Process Optimization: The goal of this step is to optimize the models to achieve the best predictive performance. This involves hyperparameter adjustments and the use of models that test strategies such as Reinforcement Learning, Ensemble Technique, and Stacking.

[0040] Process evaluation: The goal of this step is to measure whether the models actually work outside the training environment. It ensures that the predictions are reliable and accurate. Performance metrics such as classification and regression analysis, calculation of metrics, accuracy, precision, and confusion matrix are performed.

[0041] The IAP-n Process utilizes the following steps: Data Processing and Exploration, AI Model Preparation, Process Model Application, Enhanced Process Optimization, Process Evaluation, and Process Outcome, to identify crime prevention methods over time. In addition to the steps described, the model preparation stage includes Time Series analysis, including decompositions and statistical metrics on stationary series, and Cross-Validation analysis. Graphs are plotted to analyze behavioral patterns. Input: Prediction_Start_Date, Prediction_End_Date, Crime Date, Crime, Crime Prevention, and other information. The data is analyzed to identify preventive methods that can be enhanced. Techniques and Metrics: Principal Component Analysis (PCA), Auto-ARIMA, XGBoost (extreme Gradient Boosting). Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. The percentage of data used in training was 70% and in testing was 30%. Metrics: RMSE and MAE. Output: Information about the crime and the preventive methods that need to be strengthened. Example: Bird trafficking crime. Preventive methods: Environmental education for students, environmental education in communities, awareness campaigns on reporting crimes.

[0042] The IAP-E Process utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to generate knowledge for the Ethics Team formed by Artificial Intelligence so that it can respond to the demands of other processes. Entry: Country Legislation, Summaries, Decrees, Opinions, Jurisprudence, Precedents, Regulations, Treaties, Standards, International Treaties, Public Government Documents, Doctrines, Scientific Research, Specific Criminal Laws, Code of Ethics, Statute of Lawyers, Resolutions and Complementary Norms, Legislation on Human Rights, Doctrines on Psychology and Neuroscience, Complementary Databases. AI Inference Processing: In addition to the steps described, Information Retrieval (IR) techniques are applied to search, filter, and organize relevant information in large volumes of data; generative neural networks are used to create diverse scenarios, new data, and information; Large Language Models (LLMs) are used to interpret natural language; Transformers are used as a support architecture for LLMs for contextual analysis; natural language processing (NLP) techniques are used for extracting, normalizing, and interpreting information from legislation; autoregressive models, Supervised Machine Learning with Deep Learning, Recurrent Neural Networks, and a multi-agent recommendation system are configured to integrate information from the teams and propose decisions. Output: Legal information transformed into knowledge. AI Ethics team with the expertise to provide answers to the processes.

[0043] IAP-1 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to generate knowledge for the Risk Team formed by Artificial Intelligence so that it can respond to requests from other processes. In addition to the steps described, in the Data Processing / Exploration and AI Model Preparation stages, Information Retrieval (IR) techniques are applied to search, filter, and organize relevant information within large volumes of data. In the Model Application stage, texts, recommendations, and scores are generated, and Fine Tuning techniques are applied. Entry: Doctrines on Risk Management, Scientific Research, Risk Certifications, Complementary Discipline. Techniques and Metrics: Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Autoregressive Models, Recurrent Neural Networks. Output: Legal information transformed into knowledge. Risk team with the expertise to provide answers to processes.

[0044] IAP-A Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to generate knowledge for the AI-powered Reviewer so that it can respond to requests from other processes. In addition to the steps described, in the Data Processing / Exploration and AI Model Preparation stages, Information Retrieval (IR) techniques are applied to search, filter, and organize relevant information within large volumes of data. In the Model Application stage, texts, recommendations, and scores are generated, and Fine Tuning techniques are applied. Entry: Doctrines on Legal Opinions, Scientific Research on Opinions, Complementary Discipline. Techniques and Metrics: Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers. Output: Information on legislation transformed into knowledge. A team of experts with the expertise to provide answers to the processes.

[0045] The IAP-F1 process utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to identify discriminatory biases in the initial database. For discriminatory bias analysis, the Fairness techniques to be applied during the model application stage are defined in the Model Preparation stage. The enhanced optimization stage also includes analysis of Fairness metrics. The IAP-F1 process result displays anonymized data (if any).

[0046] Example Data Source: Case Study 1 https: / / dadosabertos.ibama.gov.br / dataset / Input: NOM_PESSOA_APREENSAO (Person's Name), CPF_CNPJ_PESSOA_APREENSAO (Person's CPF or CNPJ), NUM PROCESSO. Techniques and Metrics: Fairness-Aware Machine Learning Technique, Natural Language Processing, Anomanization by Suppression. Disparate Impact Ratio (DIR) Metric, Accuracy, Precision, and Confusion Matrix. The DIR technique result was <=8, indicating discriminatory bias in the process. Therefore, an anonymization technique was applied for suppression. The additional step carried out by the Ethics Committee IA ensures the effectiveness of the result. Output: NOM_PESSOA_APREENSAO (Person's Name), CPF_CNPJ_PESSOA_APREENSAO (Person's CPF or CNPJ), NUM PROCESSO anonymized and open border.

[0047] The IAP-F2 process utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to identify discriminatory biases after generating preventive methods using Generative Artificial Intelligence. For discriminatory bias analysis, the Fairness techniques to be applied during the model application stage are defined in the Model Preparation stage. The enhanced optimization stage also includes analysis of Fairness metrics. The IAP-F2 process result displays anonymized data (if any). Input: New Preventive Methods created. Example: Feasibility study for the development of animal feather sensors to be used in airports, identifying occurrences of psittacosis and zoonoses in health units to investigate possible centers of wildlife trafficking, among others. Techniques and Metrics: Fairness-Aware Machine Learning technique, Natural Language Processing, Anonymization by Suppression. Disparate Impact Ratio (DIR) metric, Accuracy, Precision, and Confusion Matrix. The Disparate Impact Ratio (DIR) metric was applied and no discriminatory bias was identified. Output: Example: There is no discriminatory bias.

[0048] The IAP-P process utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Outcome, to identify preventive methods that need to be enhanced. In addition to the steps described, the Model Application stage performs prediction and personalized recommendations, and simulates predictive scenarios. Preventive methods obtained from IAP-n are analyzed together with the prevention database. The process outcome identifies the preventive methods that need to be enhanced. Entry points: Date of the Crime, Crime, Current prevention methods, Criminal history, and other information. Examples: Environmental education for students, Environmental education in communities, Awareness campaigns about reporting crimes. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Preventive methods for enhancement. Example: Environmental education for students and environmental education in communities.

[0049] IAP-P1 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to create preventive methods using Generative Artificial Intelligence. In addition to the steps described, the Model Application stage performs prediction and personalized recommendations, and simulations of predictive scenarios. Input: Preventive methods that have been created. Example: Feasibility study for the development of animal feather sensors to be used in airports, identifying occurrences of psittacosis and zoonoses in health units to investigate possible centers of wildlife trafficking, among others. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: New preventive methods created. Example: Development of spectroscopy-based sensors to be installed in areas along wildlife trafficking routes.

[0050] The IAP-L Process utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to generate knowledge about the legislation that will be shared with the Ethics team formed by Artificial Intelligence. In addition to the steps described, in the Data Processing / Exploration and AI Model Preparation stages, Information Retrieval (IR) techniques are applied to search, filter, and organize relevant information within large volumes of data. In the Model Application stage, texts are generated, recommendations and scores are created, and fine-tuning techniques are applied. The additional optimization technique will bring greater efficiency. Entry: National legislation, international treaties, international standards, international law, case law, case summaries, opinions, scientific research, among others. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Information on legislation transformed into knowledge. Ethics team equipped to provide answers to processes.

[0051] The IAP-DH Process utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to generate knowledge about Human Rights Legislation that will be shared with the Ethics team formed by Artificial Intelligence. In addition to the steps described, in the Data Processing / Exploration and AI Model Preparation stages, Information Retrieval (IR) techniques are applied to search, filter, and organize relevant information within large volumes of data. In the Model Application stage, texts are generated, recommendations and scores are created, and Fine Tuning techniques are applied. Entry: Universal Declaration of Human Rights, International Covenant on Civil and Political Rights, Decree 98.386 which promulgates the Inter-American Convention to Prevent and Punish Torture, Decree 40 which promulgates the Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment, among other legislation. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Information on Human Rights transformed into knowledge. Ethics team equipped to provide answers to processes.

[0052] IAP-C Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to verify if preventive methods violate the laws. Corrective actions will be taken and recommendations will be made. In addition to the steps described... Entry: Legislation and Preventive Methods Techniques: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. The technical report is generated from the knowledge produced. Output: Technical report generated by the AI ​​reviewer with historical analysis, constitutional legal analysis, and the reviewer's vote with the result on approval or rejection.

[0053] IAP-C1 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to generate the risk plan. Corrective actions will be taken and recommendations will be made. In addition to the steps described, the risk classifications are defined in the model preparation stage. In the model application stage, texts are generated, recommendations and scores are created, and fine-tuning techniques are applied. In the process evaluation stage, Regulatory Risk, Confidence Score, and Explainability Index calculations are also performed. The result of this process is the risk plan. Input: Approved Preventive Methods. Example: Environmental education for students and environmental education in communities. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Risk Description, Probability, Impact, Risk Response, Action, Responsible Party, Start Date, and End Date. Example: Risk Description: Overload in handling complaints. Probability=2, Impact=2, Final Risk Assessment=2, Action: Increase call center capacity to serve the population. Responsible Party: To be defined, Start Date: To be defined, End Date: 12 / 21 / 2025, Status: Not started.

[0054] IAP-C2 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to identify new risks. In addition to the steps described, the risk classifications are defined in the model preparation stage. The model application stage generates texts, recommendations, scores, and fine-tuning techniques. The process evaluation stage also performs Regulatory Risk, Confidence Score, and Explainability Index calculations. The result of this process is the generation of new risks in the risk plan. Input: Preventive Methods and Risks. Example: Approved Preventive Methods. Example: Environmental education for students and environmental education in communities. Risk description: Overburdened in handling complaints. Probability=2, Impact=2, Final risk assessment=2, Action: Increase call center capacity to serve the population. Responsible: To be defined, Start date: To be defined, End date: 12 / 21 / 2025, Status: Not started. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks Output: New Risks Identified by Generative Artificial Intelligence. Example: Deformity in Birds Raised in Breeding Centers. Probability = 3, Impact = 3, Final Risk Assessment = 3. Action: Elimination of the Preventive Method.

[0055] IAP-M1 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to verify if there have been updates in the legislation. In addition to the steps described, the model application stage generates texts, recommendations, scores, and fine-tuning techniques. Finally, the process reports whether there have been updates to the legislation. Entry: Legislation: Universal Declaration of Human Rights, International Covenant on Civil and Political Rights, and Penal Code. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Information on legislation transformed into knowledge. Ethics team equipped to provide answers to processes.

[0056] IAP-M2 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to verify if the crime remains classified as a crime. In addition to the steps described, the Model Application stage generates texts, recommendations, scores, and fine-tuning techniques. Input: Legislation and the crime. Example: Wildlife Trafficking. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Result: Crime classified.

[0057] IAP-M3 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to verify if the preventive methods remain valid. In addition to the steps described, the Model Application stage involves generating texts, recommendations, scores, and applying Fine Tuning techniques. Entry: Legislation and preventive methods. Example: Legislation: Universal Declaration of Human Rights, International Covenant on Civil and Political Rights, Decree 98.386 promulgating the Inter-American Convention to Prevent and Punish Torture, Decree 40 promulgating the Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment. Preventive methods: Feasibility study for the development of animal feather sensors for use in airports, identifying occurrences of psittacosis and zoonoses in health units to investigate possible centers of wildlife trafficking. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Result: Valid preventive methods.

[0058] IAP-M4 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to verify if new risks have been identified after the legislation update. In addition to the steps described, the Model Application stage generates texts, recommendations, scores, and Fine Tuning techniques. Input: Risks. Example: Risk Description, Probability, Impact, Risk Response, Action, Responsible Party, Start Date, and End Date. Example: Risk Description: Overload in handling complaints. Probability = 2, Impact = 2, Final Risk Assessment = 2, Action: Increase call center capacity to serve the population. Responsible Party: To be defined, Start Date: To be defined, End Date: 12 / 21 / 2025, Status: Not started. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Risk Plan Update.

[0059] IAP-M5 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to verify Auditability compliance. In addition to the steps described, the Model Application stage involves generating texts, recommendations, scores, and applying Fine Tuning techniques. Input: Process documentation. Example: Documentation on process IAP-P1 detailing the steps performed. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Result: Audit with Corrections, Recommendations, and Best Practices.

[0060] The IAP-M6 process utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to verify algorithmic performance. In addition to the steps described, the Model Application stage involves generating texts, recommendations, scores, and applying Fine Tuning techniques. Input: Algorithms, i.e., source code of the processes. Techniques and Metrics: Asymptotic analysis with metrics, Big 0 notations, code profiling, accuracy, precision, recall, and F1 score. Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. For algorithmic performance analysis, techniques such as Bi-O Notation, Code Profiling, Accuracy, Precision, and F1Score Recall were applied. Output: Linear Search Result: 0.010836 seconds, Binary Search: 0.000011 seconds, Accuracy: 0.87, Precision: 0.8596, Recall: 0.8941 and F1 Score: 0.8765. Audit: approved.

[0061] IAP-M7 Process: Utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Outcome, to monitor the increase or decrease in crime after the application of preventive methods. In addition to the steps described, the Model Application stage generates texts, recommendations, scores, and fine-tuning techniques. Input: Updated data on crime and its prevention methods. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: Result regarding reduction or increase in crime.

[0062] The IAP-R process utilizes the following steps: Data Processing and Exploration, Preparation of AI Process Models, Application of Process Models, Enhanced Process Optimization, Process Evaluation, and Process Result, to update and generate new reports. Input: Request for a new report. Example: Risk AI Team requests a new report with detailed information for the risk. Techniques and Metrics: Natural Language Processing, Supervised Machine Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Output: New risk report generated.

[0063] Database: SQLite and MySQL were used. However, depending on the volume, Oracle, SQL Server, or other databases may be used.

[0064] SUMMARY of the main processes: The lAPs process is composed of 11 main processes and 12 subprocesses. The main processes are:

[0065] IAP-E - IA Ethics Committee

[0066] IAP-I - Risk Management Team

[0067] IAP-A - IA Reviewer

[0068] IAP-F - Preventive Artificial Intelligence - Frontiers

[0069] IAP-n - Preventive Artificial Intelligence - Crimes by criminal line with respective preventive methods.

[0070] IAP-P - Preventive Artificial Intelligence - Preventive Methods

[0071] IAP-L - Preventive Artificial Intelligence - Legislation and Legal Framework

[0072] IAP-DH - Preventive Artificial Intelligence - Human Rights

[0073] IAP-C - Preventive Artificial Intelligence - Compliance

[0074] IAP-M - Preventive Artificial Intelligence - Monitoring and Control

[0075] IAP-R - Preventive Artificial Intelligence - Reports.

[0076] The IAP-E, IAP-I, and IAP-A processes are those that form the IA teams that work on the IAPs processes. They are composed of: IA Ethics Committee (IAP-E), IA Risk Team (IAP-I), and IA Reviewer (IAP-A).

[0077] The IAP-F process is the process known as the Boundaries process. It is responsible for analyzing whether the data contains discriminatory biases, with the goal of addressing them and submitting the report for analysis and approval by the IA Ethics Committee. It consists of two boundaries, called IAP-F1 and IAP-F2.

[0078] The IAP-n process is responsible for using Artificial Intelligence to predict possible methods that will prevent crimes, thereby increasing the efficiency of the results.

[0079] The IAP-P process is responsible for linking preventive actions to the crime under analysis. Through Artificial Intelligence, it will identify which preventive methods need to be strengthened and will create new preventive methods. It has a subprocess called IAP-P1.

[0080] The IAP-L process is responsible for encompassing the country's legal system. As a local example, the IAP-L will encompass the Brazilian legal system, which will be used for crime analysis.

[0081] The IAP-DH process is responsible for encompassing legislation and treaties related to human rights.

[0082] The IAP-C process is responsible for compliance. This process performs risk analysis and also verifies whether there are violations of the legal framework belonging to the IAP-L and IAP-DH processes. It is composed of 2 subprocesses: IAP-C1 and IAP-C2.

[0083] The IAP-M process is responsible for the continuous monitoring and control of IAP processes. It is divided into 7 subprocesses: IAP-M1, IAP-M2, IAP-M3, IAP-M4, IAP-M5, IAP-M6, and IAP-M7.

[0084] The IAP-R process is responsible for managing reports generated throughout all processes and creating reports requested by the Artificial Intelligence teams. There are 19 pre-defined reports, and new reports can be generated as needed. This process is part of our responsibility to maintain transparency.

[0085] In this document, the IA Ethics Committee, IA Risk Team, and IA Reviewer will also be referred to as the Ethics Committee, Risk Team, and Reviewer.

[0086] SUMMARY of the invention: Broadly speaking, the lAPs process comprises:

[0087] I: The IAP-F process begins its border operation IAP-F1. The Virtual Border obtains the crime database(s) with its preventive methods, which will be analyzed and checked through Artificial Intelligence for discriminatory biases. If any are found, the data is processed and sent for approval by the Ethics Committee. If approved, the border is opened, and the data is sent to the next process: IAP-n. If not approved, the border will remain closed, and the data will be sent back to the origin for re-analysis and processing. The process will restart when new data is submitted.

[0088] II: The IAP-n process obtains data from the IAP-F1 process, uses Artificial Intelligence to make predictions based on the application of preventive methods according to the Criminal Guidelines, and sends the report for approval by the Ethics Committee. If the report is approved, the data is sent to the subsequent IAP-P, IAP-P1, and IAP-R processes. If the report is not approved, new Artificial Intelligence inferences may be performed and a report sent to the Ethics Committee. The AI ​​Ethics Committee may update the report for approval.

[0089] III: The IAP-P process obtains data from the IAP-n process and relates preventive methods for Crimes in the Criminal Line. In this way, through Artificial Intelligence, the process can identify which preventive methods need to be enhanced. These methods to be enhanced are sent to the Ethics Committee, which will approve or reject the selected methods. In addition to identifying the methods that need to be enhanced, through the IAP-P1 subprocess, it will be possible, through Artificial Intelligence, to identify new preventive methods. The IAP-P1 process needs to pass through the second frontier, IAP-F2, to verify that the new preventive methods generated by Generative Artificial Intelligence do not have discriminatory biases. After passing through the IAP-F2 frontier, the data will be sent to the Compliance process, IAP-C.

[0090] IV) The IAP-L process encompasses the country's legislation and legal framework. Knowledge is generated through Artificial Intelligence.

[0091] V) The IAP-DH process encompasses Human Rights Legislation. Knowledge is generated through Artificial Intelligence.

[0092] VI) The IAP-C process obtains information about the IAP-L, IAP-DH, IAP-P, IAP-P1, and IAP-F2 processes; that is, it verifies whether the preventive methods violate the laws. Corrective actions will be taken and recommendations will be made. The Ethics Committee is responsible for approving or rejecting the final report on the preventive methods. After approval, the Risk Analysis begins.

[0093] VII) The IAP-C1 and IAP-C2 subprocesses are responsible for performing the Risk Analysis.

[0094] VIII) The IAP-M1 process performs Monitoring and Control over updates to the Legislation that are part of the IAP-L and IAP-DH processes.

[0095] IX) The IAP-M2 process is responsible for verifying whether, after updating the legislation, the crime in question continues to be classified as such under the legislation.

[0096] X) The IAP-M3 process uses Artificial Intelligence to verify whether, after the legislation update, the preventive methods remain valid because they do not violate the laws.

[0097] XI) The IAP-M4 process uses Artificial Intelligence to verify if there are new risks resulting from updated legislation, and also performs Risk Monitoring.

[0098] XII) The IAP-M5 process uses Artificial Intelligence to verify whether the information relevant for auditability is in compliance.

[0099] XIII) The IAP-M6 process performs algorithmic management, analyzing the performance and transparency of the algorithms.

[0100] XIV) The IAP-M7 process monitors the decrease or increase in crime after the implementation of preventive methods.

[0101] XV) The IAP-R process manages the reports that are generated throughout all IAP processes.

[0102] Brief description of the drawings

[0103] Figure 1 presents the flowchart with an overview of the integration of the 23 IAP processes: IAP-E, IAP-I, IAP-A, IAP-F, IAP-F1, IAP-F2, IAP-n, IAP-DH, IAP-L, IAP-P, IAP-P1, IAP-C, IAP-C1, IAP-C2, IAP-M, IAP-M1, IAP-M2, IAP-M3, IAP-M4, IAP-M5, IAP-M6, IAP-M7 and IAP-R.

[0104] Figure 2 presents the flowchart of the IAP-E process.

[0105] Figure 3 presents the flowchart of the IAP-I process.

[0106] Figure 4 presents the flowchart of the IAP-A process.

[0107] Figure 5 presents the flowchart of the IAP-F1 process.

[0108] Figure 6 presents the flowchart of the IAP-F2 process.

[0109] Figure 7 presents the flowchart of the IAP-n process.

[0110] Figure 8 presents the flowchart of the IAP-P process.

[0111] Figure 9 presents the flowchart of the IAP-P1 process.

[0112] Figure 10 presents the flowchart of the IAP-L process.

[0113] Figure 11 presents the flowchart of the IAP-DH process.

[0114] Figure 12 presents the flowchart of the IAP-C process.

[0115] Figure 13 presents the flowchart of the IAP-C1 process.

[0116] Figure 14 presents the flowchart of the IAP-C2 process.

[0117] Figure 15 presents the flowchart of the IAP-M1 process.

[0118] Figure 16 presents the flowchart of the IAP-M2 process.

[0119] Figure 17 presents the flowchart of the IAP-M3 process.

[0120] Figure 18 presents the flowchart of the IAP-M4 process.

[0121] Figure 19 presents the flowchart of the IAP-M5 process.

[0122] Figure 20 presents the flowchart of the IAP-M6 process.

[0123] Figure 21 presents the flowchart of the IAP-M7 process.

[0124] Figure 22 presents the flowchart of the IAP-R process.

[0125] Description of the invention

[0126] To ensure clear understanding, each step of the process will be detailed: an overview of the process, process prerequisites (if any), the role of AI teams in the respective process, information on reports, and a detailed flowchart.

[0127] The flowcharts will represent the steps involved in the processes. Each rectangular box represents an activity. The arrows represent the flow of data or information to other activities. The cylinders represent storage media. The actions may be sequential or not. It is important to note that the flowchart has been numbered, and this numbering is not sequential. This numbering is a reference to facilitate understanding of the logical flow. The defined processes will be understood by professionals with expertise in the Legal and Technological fields.

[0128] The AI ​​inferences are in flowcharts 1400 to 1410 and 1500 to 1509.

[0129] IAP-E - IA Ethics Committee

[0130] Overview: The Ethics Committee is formed using knowledge generated by Artificial Intelligence. They are responsible for "approving or not" reports and process results. They operate at the borders and in other LAP (Laboratory Action Planning) processes. Their knowledge base is composed of Law, Human Rights, Risk, History, Sociology, Psychology, and Neuroscience. They are able to acquire knowledge in other disciplines. The knowledge is generated from the following databases:

[0131] Database 1 - National Legislation, Legal System: Contains the country's legislation, summaries, decrees, opinions, case law, precedents, and regulations. Database 2 - Treaties, Norms and International Legislation: Includes the International Treaties to which the country is a party. Database 3 - Public Government Documents: Includes Public Government Documents. Database 4 - Doctrines: Includes Doctrines. Database 5 - Scientific Research: Includes Scientific Research. Database 6 - Selected Legal Framework: Includes the selected legal framework as analyzed on screen. This database is expanded according to the needs of each crime. For example, if there is a specific law regarding the crime being analyzed that was not part of Database 1, that law will be included in this database. For example: LAW No. 14,994, OF OCTOBER 9, 2024 (Amends Decree-Law No. 2,848, of December 7, 1940 (Penal Code), Decree-Law No. 3,688, of October 3, 1941 (Law on Criminal Offenses), Law No. 7,210, of July 11, 1984 (Law on Criminal Execution), Law No. 8,072, of July 25, 1990 (Law on Heinous Crimes), Law No. 11,340, of August 7, 2006 (Maria da Penha Law) and Decree-Law No. 3.Law 689 of October 3, 1941 (Code of Criminal Procedure), to make femicide an autonomous crime, to increase its penalty and that of other crimes committed against women for reasons of their gender, as well as to establish other measures aimed at preventing and curbing violence against women. Database 7 - Codes of Ethics. Database 8 - Statute of Lawyers. In the case of Brazil, it is the Statute of the Brazilian Bar Association. Database 9 - Resolutions and Complementary Standards. Database 10 - Legislation on Human Rights. Database 11 - Doctrines and Scientific Research on History. Database 12 - Doctrines and Scientific Research on Sociology. Database 13 - Doctrines and Scientific Research on Psychology and Neuroscience. Database 14 - Doctrines and Scientific Research - Complementary Discipline. Important disciplines for the Ethics Committee's knowledge base can be identified. Depending on the type of crime analyzed, the Ethics Committee may need to acquire new knowledge, that is, new content. For example: For the analysis of the crime of racism, it may be necessary to include Doctrines and Scientific Research on Social Sciences. In this case, this information will be included in Database 14. In this way, information that was not included in the 13 previous databases, but which is identified as important for the LAPs processes, will be included in this database.

[0132] The number of databases is not limited to those described above. New databases may be added to the Ethics Committee.

[0133] They use AI inference to make decisions.

[0134] Prerequisites: The databases must be obtained from secure and reliable sources.

[0135] Report: No report was generated for this process.

[0136] Detailed Logical Flow of Figure 2: The IAP-E process begins by obtaining information from the databases (1100). Through IA Inference, a knowledge base is generated (1101). The objective is to generate knowledge so that the IA Ethics Committee has the necessary information to make decisions throughout the IAP processes.

[0137] IAP-1 - Risk IA Team

[0138] Overview: Process formed by knowledge obtained from the Ethics Committee and by 4 risk-related databases: Database 1 - Doctrines on Risk Management. Database 2 - Scientific Research on Risks. Database 3 - Risk Certifications. Database 4 - Complementary Discipline. Depending on the type of crime analyzed, the Risk AI team may need to acquire new knowledge in a new discipline. It uses AI inference to make decisions.

[0139] Prerequisites: The databases must be obtained from secure and reliable sources.

[0140] Report: No report was generated for this process.

[0141] Detailed Logical Flow Figure 3: The IAP-I process will obtain knowledge from the Ethics Committee (1200). Then it will obtain information from the databases (1201). Through AI Inference, a knowledge base is generated (1202). The objective is to generate knowledge so that the Risk AI Team has a knowledge base to make decisions.

[0142] IAP-A - IA Reviewer

[0143] Overview: The Reviewer is responsible for analyzing violations of the law. Guided by the Sovereign Constitution, they seek to unequivocally maintain democracy by invalidating actions that violate legislation. The process is based on knowledge obtained from the Ethics Committee, the Risk Team, and three databases: Database 1 - Doctrines on Legal Opinions. Database 2 - Scientific Research on Opinions Database 3 - Complementary Discipline. Depending on the type of crime analyzed, the reviewer may need to acquire new knowledge in a new discipline.

[0144] It uses AI inference to make decisions.

[0145] The number of databases is not limited to those described above. New databases may be added to the Reviewer.

[0146] Prerequisites: The databases must be obtained from secure and reliable sources.

[0147] Report: No report was generated for this process.

[0148] Detailed Logical Flow Figure 4: The IAP-A process will obtain the knowledge base from the Ethics Committee (1300). It will obtain the knowledge base from the Risk Team (1301). Then it will obtain information from the databases (1302). Through AI Inference, knowledge bases are generated. (1303) The objective is to generate a knowledge base so that the Reviewer has the knowledge to generate the technical opinion, approve or not the result.

[0149] IAP-F - Preventive Artificial Intelligence - Frontiers

[0150] Overview: Preventing discrimination is fundamental in LAP processes. Therefore, we reinforce actions to protect against any type of direct or indirect discrimination. It is essential to prevent technology from developing a new form of prejudice with discriminatory biases that violate fundamental principles.

[0151] It utilizes the Discriminatory Biases Database. This database contains information such as: age, gender, race / ethnicity, nationality, sexual orientation, marital status, disability, religion, education, family income, job title, and salary. The database may be updated with new data and information as needed to expand the analysis of discriminatory biases.

[0152] Prerequisites: 1 - The data submitted to begin processing must comply with the General Data Protection Law (LGPD). This initial data is not part of the lAPs process, but rather belongs to the agency / institution / company providing the data for analysis. Although the initial data is not part of the lAPs process, it will be analyzed by the first frontier, thus initiating verification against discriminatory biases and personal identification. If not approved, lAPs will indicate which data were rejected. 2 - The initial database must be obtained from secure, reliable, and complete sources. 3 - Just as legislation evolves, so does technology and, consequently, artificial intelligence.In specific cases where sensitive data is necessary for the analysis of the crime in question, such as "race" in the crime of racism, this data cannot reference / relate to any form of personal identification, and criteria will be adopted to guarantee non-discrimination, thus maintaining equality and guaranteeing fundamental principles. 4 - The lAPs process can never be used to identify people or attribute crimes to people, and it cannot be used as a tool to help identify people. If the initial database contains data that allows for the identification of people, this data must be processed, for example, by applying anonymization techniques to guarantee privacy and ensure compliance with the LGPD (General Data Protection Law in Brazil).

[0153] When there is no discriminatory data in the database to be analyzed, the result obtained will be considered to be that there are no discriminatory biases.

[0154] Report: This process generates two reports: the IAP-R1 report and the IAP-R1A report. The IAP-R1 report contains information on discriminatory biases found in the initial data after approval by the Ethics Committee. The IAP-R1A report contains information on discriminatory biases found in the initial database that were rejected, preventing further processing in subsequent cases.

[0155] IA Team's Role: The Ethics Committee will be responsible for approving or rejecting the IAP-R1 report.

[0156] Detailed flowchart figure 5: The IAP-F process is initiated by the IAP-F1 process. The database to be analyzed by the lAPs (200) process is obtained, that is, the database of the crime in question provided by the Institution / Body / Company will be analyzed. The border is closed at this point.

[0157] Next, a critical analysis of the crime data is performed using AI Inference. This critical analysis uses the Discriminatory Biases Database to check if there is any discrimination (201).

[0158] If no discriminatory biases are identified after the IA Inference, the IAP-R1 Report will be generated and sent to the Ethics Committee for approval or rejection (203). If the Report is approved by the Ethics Committee, the boundary will be opened and the data will be made available for future steps: IAP-R, IAP-n or IAP-M7 (205). If the report is rejected by the Ethics Committee, that is, if the Ethics Committee understands that there are indeed discriminatory biases, the IAP-R1 A report will be generated containing information on the rejected data. This report will be sent to the Institution / Body / Company so that they can analyze the result, since this database did not pass through the IAP-F1 boundary (206).

[0159] If discriminatory biases are found, these biases will be addressed through AI Inference, being removed or mitigated (207). Then, the IAP-R1 Report is generated and sent to the Ethics Committee for approval or rejection (208). If the Report is approved, the boundary will be Opened and the data will be made available for future steps: IAP-R, IAP-n, or IAP-M7 (205). If the report is not approved, meaning the process identified that discriminatory biases still exist, a new AI inference may be performed (210) and the data will be sent for processing (201). This decision is made by the Ethics Committee. If the Ethics Committee decides not to perform a new AI Inference, the IAP-R1A report will be generated on these rejected data and this report will be sent to the Institution / Body / Company so that they can perform a new analysis, as this database did not pass through the boundary (206).

[0160] IAP-F2 - Preventive Artificial Intelligence - Frontiers for New Preventive Methods Generated by Artificial Intelligence

[0161] Overview: The IAP-F2 process is the second frontier. This frontier aims to verify whether new preventive methods created by Artificial Intelligence have discriminatory biases. It works similarly to the IAP-F1 process.

[0162] Prerequisites: The IAP-P1 process must have been completed, as this process generates the new preventive methods.

[0163] Reports: This process generates the IAP-R5. This report will contain information about the new preventive methods, highlighting whether or not biases were identified and how many inferences were made by Artificial Intelligence until crossing the border.

[0164] Role of the IA Team: The Ethics Committee will be responsible for approving or rejecting the IAP-R5 report.

[0165] Detailed flowchart of Figure 6: The IAP-F2 process is initiated by the IAP-P1 process (250). Information is obtained about the new preventive methods that were created by Artificial Intelligence in the IAP-P1 process. The boundary is closed at this point.

[0166] Next, a critical analysis of the new preventive methods is carried out using AI Inference (251). This analysis checks against the Discriminatory Biases Database whether there is any discrimination.

[0167] If discriminatory biases are found (252), these biases will be addressed, removed, or mitigated, and a new result will be generated (257). The IAP-R5 Report will be generated and sent to the Ethics Committee for approval (258). If the result is not approved, i.e., the process identified that discriminatory biases still exist, a new IA Inference may be performed (260). This decision is made by the Ethics Committee. If a new IA Inference is performed, the information will be forwarded for processing (251). If the Ethics Committee decides not to perform a new IA Inference, the Ethics Committee will re-analyze, adjust, and approve the IAP-R5 (256). The Border will be opened, and the information will be made available for the IAP-C and IAP-R processes (255).

[0168] If there are no discriminatory biases after the IA Inference (251), the IAP-R5 Report will be generated. The IAP-R5 Report is sent to the Ethics Committee, which may or may not approve the report (253). If the Report is approved (254), the boundary will be Opened and the data will be sent to the next steps: IAP-C and IAP-R (255). If the report is rejected, that is, if there are still discriminatory biases, the Ethics Committee must re-analyze the IAP-R5 report, keep only the approved methods and approve the IAP-R5 report (256). Then the Boundary will be opened and the information will be made available for the future IAP-C and IAP-R processes (255).

[0169] The IAP-n process - Preventive Artificial Intelligence - Crimes

[0170] Overview: The IAP-n process is responsible for using Artificial Intelligence to make predictions and apply possible methods to prevent crime, thereby increasing the efficiency of the results.

[0171] Prerequisites: For the IAP-n process to begin, the data analyzed by the IAP-F process must have passed through the border. This ensures the data is ready for analysis.

[0172] Team Role: The Ethics Committee team is responsible for approving the reports.

[0173] Detailed flowchart figure 7: The first step is to obtain the data from the IAP-F1 process (300).

[0174] After receiving the data, the Prediction period (301) will be defined.

[0175] Based on the data received, exploratory data analysis will be performed (302). Exploratory analysis is the stage where graphs are plotted to obtain an initial understanding of crime and preventive methods.

[0176] Perform AI Inference (303). Upon completion of the analysis, the IAP-R2 Report will be generated and sent to the Ethics Committee for approval or rejection (304).

[0177] When can the report be rejected? The report will be rejected when the Ethics Committee analyzes the result and understands that the Prediction was based on discriminatory information. If the report is rejected, the Ethics Committee may decide whether a new IA Inference will be performed (306). If a new inference is performed, the data will go through the process again (303) and the IAP-R2 report will be updated so that it can be sent again by the Ethics Committee (304). This procedure will be performed as many times as requested by the Ethics Committee. If the Ethics Committee chooses not to perform a new IA Inference, the Committee will re-analyze, adjust and approve the Report (307). After the report is approved, the information will be made available for future processes: IAP-P, IAP-P1 and IAP-R (308).

[0178] The IAP-L process - Preventive Artificial Intelligence - Legislation

[0179] Overview: The IAP-L process encompasses the country's legislation, standards, case law, case summaries, opinions, international treaties to which the country is a party, and the entire framework that forms part of the legal system. This process is composed of 6 databases, which include: Database 1 - Country Legislation, Legal System: Contains the country's legislation, standards, case law, decrees, opinions, case law, precedents, and regulations. Database 2 - Treaties, Norms and International Legislation: Includes the International Treaties to which the country is a party. Database 3 - Public Government Documents: Includes Public Government Documents. Database 4 - Doctrines: Covers the Doctrines. Database 5 - Scientific Research: Covers scientific research. Database 6 - Selected Legal Framework: Includes the selected legal framework as analyzed on screen. This database is expanded as needed for each crime.

[0180] Prerequisites: The databases must be obtained from secure and reliable sources.

[0181] AI Team Involvement: In this process, there is no involvement from the teams.

[0182] Reports: This process generates the IAP-R6 report. This report presents information about the legislation, such as the effective date and updates.

[0183] Logical Flow Detail Figure 10: The IAP-L process is initiated by obtaining information from the databases (500). Through AI Inference, a knowledge base is generated (501). The objective is to generate a knowledge base so that other processes can perform queries and verify if there is a violation of these laws. After generating the knowledge base, the IAP-R6 Report will be generated (502). Then the information will be made available to the IAP-C and IAP-M1 processes (503).

[0184] The IAP-DH process - Preventive Artificial Intelligence - Human Rights

[0185] Overview: The IAP-DH process encompasses the legal framework related to Human Rights. Examples of Human Rights legislation include: the Universal Declaration of Human Rights, the International Covenant on Civil and Political Rights, Decree 98.386 which promulgates the Inter-American Convention to Prevent and Punish Torture, Decree 40 which promulgates the Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment, among other legislation.

[0186] Prerequisites: The database must be obtained from secure and reliable sources.

[0187] AI Team Involvement: In this process, there is no involvement from the teams.

[0188] Reports: This process generates the IAP-R7 report. This report presents information about the relevant legislation.

[0189] Detailed Logical Flow Figure 11: The IAP-DH process is initiated by obtaining information from the Human Rights Legislation Database (400). Through AI Inference, a knowledge base is generated (401). The objective is to generate a knowledge base so that other processes can perform queries and verify if there is a violation of these Laws. After generating the knowledge base, the IAP-R7 Report will be generated, which will contain information about the Legislation (402), such as the effective date and update date. Then, the information will be made available to the IAP-C and IAP-M1 processes (403).

[0190] The IAP-P process - Preventive Artificial Intelligence - Prevention

[0191] Overview: The objective of this process is to identify and / or create preventive methods that need to be strengthened or developed. These methods will be connected to the areas of Education, Health, Poverty Eradication, Security, among others. The IAP-P has the subprocess IAP-P1.

[0192] The IAP-P1 creates new preventive methods. It has a higher inference capability, analyzing a pattern of behavior to suggest new preventive methods by identifying iterations that are not so obvious. The preventive methods are ordered according to their predicted effectiveness.

[0193] Prerequisites: This process can only begin after the IAP-n file has been processed.

[0194] IA Team's Role: The Ethics Committee will be responsible for analyzing and approving or rejecting the IAP-R3 report. The main focus is to assess whether the preventive methods violate Human Rights.

[0195] Reports: This process generates the IAP-R3 Report. This report contains information about preventive methods.

[0196] Logical flow detail figure 8: Obtain information from process IAP-n (600). Through AI Inference, process IAP-P will, based on the preventive methods already defined in the database, relate the crime to these methods (601). The result of the analysis performed by process 601 will be sent to process IAP-P1 (602).

[0197] Identify which preventive methods need to be strengthened to prevent this crime from occurring in the future (603).

[0198] The IAP-R3 Report will be generated. From then on, this report will be sent to the Ethics Committee for approval or rejection (604).

[0199] If the report is approved by the Ethics Committee, this information will be made available for the IAP-C and IAP-R processes (607).

[0200] If the report is not approved, the Ethics Committee will re-analyze, adjust and approve it (606), thus proceeding to the next stage which is to make information available for the IAP-C and IAP-R processes (607).

[0201] The IAP-P1 process - Preventive Artificial Intelligence - Prevention with new Preventive Methods

[0202] Overview: This process uses Artificial Intelligence to create new preventive methods. This process will use approved preventive methods as a reference for learning and obtaining new guidelines. Multifaceted preventive actions can be identified by AI. The following are prohibited in IAP-P1 preventive methods: preventive development related to any type of sanction, deprivation of liberty, penalty, imprisonment, restriction or limitation of liberty, internment, and disrespect, abuse, or violation of human rights.

[0203] Prerequisites: This process can only begin after the IAP-n and IAP-P files have been processed.

[0204] IA Team's Role: The Ethics Committee will be responsible for analyzing and approving or rejecting the IAP-R4 report. The main focus is to assess whether the new preventive methods violate Human Rights and whether they are aligned with the country's legislation. The Ethics Committee will play a fundamental role, as it needs to conduct an in-depth analysis of these methods, and this will be crucial for their inclusion or exclusion in the IAP-R4 Report.

[0205] Report: This process will generate the IAP-R4 Report.

[0206] Logical Flow Detail Figure 9: The process begins by obtaining information from processes IAP-n and IAP-P (620). Through AI Inference, process IAP-P1, based on the already defined preventive methods, will relate the crime to these methods and generate new preventive methods (621). Report IAP-R4 will be generated with information on the new preventive methods (622) (if any). From then on, this report will be sent for approval by the Ethics Committee. If the report is approved, this information will be made available to processes IAP-C and IAP-R (625). If the report is not approved, the Ethics Committee will re-analyze, adjust, approve or not the selected methods (624). The new methods created that are rejected will be deleted. After approval of the report, the information will be made available to processes IAP-C and IAP-R (625).

[0207] IAP-C - Preventive Artificial Intelligence - Compliance

[0208] Overview: The IAP-C process is responsible for verifying whether preventive methods comply with legislation and the entire legal framework. Through AI inference, it detects whether or not there is a violation of the country's legislation. In this way, it verifies compliance with the entire legal framework that is part of the processes and identifies which actions may or may not succeed. In the case of Brazil, it is of fundamental importance, as a democratic state governed by the rule of law and constituted by the "fundamental clauses" of the Constitution, that preventive methods do not violate the Constitution, thus maintaining the integrity of its objectives. Furthermore, the technical opinion prepared by the AI ​​reviewer is necessary in this process. The technical opinion must contain a legal-constitutional analysis of the IAP-R8 report.

[0209] Prerequisites: The IAP-L, IAP-DH, IAP-P, IAP-P1, and IAP-F2 processes must have been completed.

[0210] IA Team's Role: The Ethics Committee will be responsible for approving or rejecting the IAP-R8 report. The IA Reviewer will prepare the IAP-R9 Constitutional Technical Opinion on preventive methods and will be responsible for approving or rejecting the opinion's findings.

[0211] Reports: This process generates the IAP-R8 and IAP-R9 reports.

[0212] The IAP-R8 report will contain information regarding the non-violation of legislation. The IAP-R9 report will contain the Legal Opinion.

[0213] Logical Flow Detail Figure 12: The IAP-C process will obtain information from the processes IAP-L, IAP-DH, IAP-P, IAP-P1, and IAP-F2; that is, this process will access the knowledge base of Legislation, the knowledge base on Human Rights, information on preventive methods, and on the liberation of the second frontier (700). It will be verified through AI Inference whether the preventive methods violate any Legislation of the Legal Framework (701). If there is a violation (702), corrective actions will be implemented (703). If there is no violation of the laws, the IAP-R8 Report will be generated and sent for approval or not by the Ethics Committee (704). If the report is approved, the information will be made available to the AI ​​Reviewer who will prepare the IAP-R9 Technical Opinion (708), and approve or not it. If IAP-R8 is not approved, the Ethics Committee will decide whether further corrections will be implemented (706) or whether the Committee itself will adjust the report for approval.If new corrections are implemented, process (703) will be carried out; otherwise, the report will be re-analyzed, adjusted, and approved by the Ethics Committee (707). The IAP-R8 report will be sent so that the Reviewer can prepare the Technical Opinion (IAP-R9) (708) and approve or not the result. The Technical Opinion is generated based on the IAP-R8 Report.

[0214] If the result of the opinion is approved by the IA Reviewer, the information will be made available for processes IAP-C1, IAP-C2 and IAP-R (711). If the result of the opinion is not approved by the Reviewer, IAP-R9 will be sent to the Ethics Committee which will adjust IAP-R8 according to the opinion (710) and the information will be made available for processes IAP-C1, IAP-C2 and IAP-R (711).

[0215] IAP-C1: Preventive Artificial Intelligence - Risk Analysis

[0216] Overview: This process aims to manage risk. To achieve this, it utilizes a Risk Database. This database is regularly updated by the Ethics Committee; that is, every time the Ethics Committee identifies a new risk, approves or rejects the report, this Risk Database is updated.

[0217] The values ​​for the probability of the risk occurring, the impact the risk will cause, and the criticality of the risk may vary according to the complexity of the crime analyzed as well as the complexity of the data provided by the agency / institution / company.

[0218] Prerequisites: The IAP-C process must have been completed.

[0219] IA Team Role: The Risk Team will be responsible for approving or rejecting the IAP-R10 report. In addition to Risk Team approval, the report must also be approved by the Ethics Committee.

[0220] Reports: This process generates the IAP-R10 Report, which contains the Risk Plan.

[0221] Logical flow breakdown figure 13: The IAP-C1 process will obtain the information from the IAP-C process (720). Through AI Inference, it will perform the Risk analysis (721). After this analysis, the IAP-R10 Report will be generated with the risk plan. This report will be sent to the Risk Team, which will approve or reject the report (722).

[0222] If the report is approved, Risk Treatment Through Inference AI (726) must be performed. After the Risks have been properly treated, the risk plan will be updated and sent for approval or not by the Ethics Committee (727). If the risk plan is approved by the Ethics Committee, the information will be made available to IAP-M4, IAP-R and IAP-C2 (730). If the Risk plan is not approved by the Ethics Committee, IAP-R10 will be adjusted, updated and approved (729) by the Ethics Committee. Then the information will be made available to IAP-M4, IAP-R and IAP-C2 (730).

[0223] If the report is not approved (723), the Risk Team will decide whether to perform a new AI Inference (724). If a new Inference is performed, it will proceed to the process (721). If a new Inference is not performed, the Risk Team will adjust, update and approve (725). Then, it will proceed to the risk treatment process (727) proceeding to the next steps.

[0224] IAP-C2: Preventive Artificial Intelligence - Identification of New Risks

[0225] Overview: This process aims to identify risks that have not yet been identified. Generative Artificial Intelligence will be used to identify new risks.

[0226] Prerequisites: The IAP-C and IAP-C1 processes must have been completed.

[0227] IA Team Role: The Risk Team will be responsible for approving or rejecting the IAP-R11 report, which contains the Risk Plan with the newly identified Risks. In addition to the Risk Team's approval, the report must also be approved by the Ethics Committee.

[0228] Reports: This process generates the IAP-R11 Report, which contains the risk plan for the new risks integrated with IAP-R10.

[0229] Logical flow breakdown figure 14: The IAP-C2 process will obtain information from the IAP-C and IAP-C1 processes (740). Through AI Inference, it will identify new Risks (741). After this step, the IAP-R11 Report will be generated with the risk plan. This report will be sent to the Risk Team, which will approve or reject the report (742).

[0230] If the report is approved (743), it will be verified whether new risks have been identified (746). If new risks are identified, Risk Treatment Through AI Inference must be performed (747). After the Risks have been properly treated, the risk plan will be updated and sent for approval by the Ethics Committee (748). If the risk plan is approved by the Ethics Committee, the information will be made available to IAP-M4 and IAP-R (751). If the Risk Plan is not approved (749), IAP-R11 will be adjusted, updated and approved by the Ethics Committee (750). Then the information will be sent to IAP-M4 and IAP-R (751). If the report is rejected, the Risk Team will decide whether a new AI Inference will be performed (744). If a new Inference is performed, it will proceed to process (741), otherwise, the IAP-R11 will be sent to the Risk Team so that they can adjust, update and approve the IAP-R11 report (745). Then, it will proceed to process (747).

[0231] If the report is not approved (723), the Risk Team will decide whether to perform a new AI Inference (744). If a new Inference is performed, it will proceed to the process (741). If a new Inference is not performed, the Risk Team will adjust, update and approve (745). Then, it will proceed to the risk treatment process (747) proceeding to the next steps.

[0232] IAP-M: Preventive Artificial Intelligence - Monitoring and Control.

[0233] This process is responsible for Monitoring and Control, that is, the tracking of IAP processes. It is a fact that Artificial Intelligence is a powerful technology, and we have no visibility into its potential future power. Therefore, developing processes with Artificial Intelligence requires responsibility and compliance with legal requirements, and especially a commitment to maintaining Transparency, Explainability, and Auditability. Thus, Monitoring and Control is essential for IAP processes. The IAP-M process is initiated by the IAP-M1 process.

[0234] IAP-M1: Preventive Artificial Intelligence - Monitoring and Control for generating a knowledge base.

[0235] Overview: This process aims to generate a knowledge base for the other Monitoring and Control processes.

[0236] Prerequisites: The IAP-L and IAP-DH processes must have been completed.

[0237] AI Team Role: There is no team involvement in this process.

[0238] Reports: No reports are generated in this process.

[0239] Logical Flow Detail Figure 15: The IAP-M1 process will obtain information from the IAP-L and IAP-DH processes (800). Through AI Inference, a knowledge base will be generated. This knowledge base will be used in the other Monitoring and Control processes (801). The information is then made available to the Ethics team (802) for the IAP-M2, IAP-M3, and IAP-M4 processes (803).

[0240] Figure 16: IAP-M2 - Preventive Artificial Intelligence - Monitoring and Control of Crime Classification

[0241] Overview: This process aims to verify whether, after an update to the legislation, the crime remains classified as such. It's possible that a particular crime may cease to be considered a crime in the future due to an update to the law. As an example, we can cite "treason," which was previously considered a crime but, due to an update to the legislation, is no longer considered one. Therefore, we can only proceed with the LAPs (Law of Administrative Proceedings) if the act committed remains classified as such.

[0242] Prerequisites: The IAP-M1 process must have been completed.

[0243] Reports: This process generates the IAP-R12 and IAP-R13 reports.

[0244] The IAP-R12 report states that the crime has been classified.

[0245] The IAP-R13 report states that the crime is not classified as a specific offense and justifies this by citing the new law.

[0246] Actions of the AI ​​Team: The Ethics Committee and Risk Team will be informed when the crime committed is no longer considered a crime after new legislation.

[0247] Logical flow detail figure 16: The IAP-M2 process will obtain the information from the IAP-M1 processes (820).

[0248] Through AI Inference, it will be verified whether the crime in question continues to be classified under the Legislation, that is, whether it continues to be considered a crime (821).

[0249] If the crime that was committed continues to be considered a crime, the IAP-R12 report (824) will be generated and the information will be made available for the IAP-M3 and IAP-R processes (825).

[0250] If the crime that was committed ceases to be considered a crime, report IAP-R13 (823) will be generated. The Ethics Committee and Risk Team will be informed of the results of these reports. In this way, there will be no continuation of the other lAPs processes.

[0251] Figure 17: IAP-M3 - Preventive Artificial Intelligence - Monitoring and Control of Preventive Methods.

[0252] Overview: This process aims to verify whether preventive methods remain valid after legislative updates. Following the law's update, some preventive methods may no longer comply with the law.

[0253] Prerequisites: The IAP-M1 process must have been completed.

[0254] Reports: This process generates the IAP-R14 report. This report indicates whether the preventive methods remain valid or not.

[0255] IA Team's Role: The Ethics Committee and Risk Team are responsible for approving or rejecting the IAP-R14 report.

[0256] Logical flow detail figure 17: The IAP-M3 process will obtain information from the IAP-M1 processes (840). Through AI Inference, it will be analyzed whether the preventive methods remain valid (841).

[0257] If preventive methods remain valid, the IAP-R14 report will be generated and sent to the Ethics Committee and Risk Team (843). The information will then be made available for the IAP-M4 and IAP-R processes (845).

[0258] If preventive methods are no longer valid, the IRP-R14 Report will be generated and sent for re-analysis, updating and approval by the Ethics Committee (844). The information will be made available for the IAP-M4 and IAP-R processes (845).

[0259] Figure 18: IAP-M4 - Preventive Artificial Intelligence - Risk Monitoring and Control

[0260] Overview: This process aims to monitor risks.

[0261] Prerequisites: The IAP-C1, IAP-C2, and IAP-M3 processes must have been completed.

[0262] Reports: This process updates the IAP-R10 and IAP-R11 reports.

[0263] Role of the AI ​​Team: The Ethics Committee and Risk Team are responsible for approving / updating or not the reports.

[0264] Logical flow detail figure 18: The IAP-M4 process will obtain information from the IAP-C1, IAP-C2 and IAP-M3 processes (860). Through AI Inference, it will be verified whether the risk plan needs to be updated (861).

[0265] If it becomes necessary to update the Risk Plan, the IAP-R10 and IAP-R11 reports will be updated. The IAP-R11 report will be sent for approval by the Ethics Committee and Risk Team (864). The IAP-R10 report does not require approval, as it is currently contained within the IAP-R11 report. If the report is approved, risk implementation and management will be carried out (867). Monitoring and Control will be performed continuously (868). If the report is not approved, the Ethics Committee must analyze, update, and approve the IAP-R11 report (866). The report information will be made available to IAP-R (863). Following this, risk implementation and management will be carried out (867). Monitoring and Control will be performed continuously (868).

[0266] If it is not necessary to update the Risk plan (862), Continuous Monitoring IAP-M4 (868) will be carried out.

[0267] IAP-M5 - Preventive Artificial Intelligence - Monitoring and Control for Auditability

[0268] Overview: This process aims to perform monitoring and control for auditability, ensuring that processes are functioning in accordance with the principles of transparency, explainability, efficiency, and effectiveness. It also aims to verify whether the implementation plan is being followed. Transparency: The origin of the database and the legal framework must be clearly explained in the documentation. Whenever there are updates that impact the processes, the documentation must be updated, keeping them properly mapped and updated in the Documentation Database. Explainability: This is the ability to explain all processes. Reports are documents that will be used for explainability. All documents must be complete. Algorithmic decisions need to be equitable and unbiased, and therefore must be justified. Effectiveness: We will analyze whether the IAP process achieved its goal of reducing the number of crimes. Efficiency: We will conduct a performance analysis by comparing results over the years. Monitoring and control of the audit will be an ongoing step. For transparency and explainability analysis, IAP-M5 will obtain information from IAP-M6. To perform the effectiveness and efficiency analysis, IAP-M5 will obtain information from IAP-M7.

[0269] Prerequisites: It is necessary that the IAP-F, IAP-P, IAP-C, IAP-L, IAP-DH, IAP-M6, and IAP-M7 processes have been processed. To be compliant, the processes must contain: Documentation on all processes including, but not limited to: process steps and expected results, database queries, and generated reports. Name and description of the databases consulted by the processes.

[0270] Reports: This process generates the IAP-R15 and IAP-R16 reports. The IAP-R15 report will present information about the audit objective and expected results, including information on the results found. When the processes are in compliance, the report will include: the audit objective, the Artificial Intelligence techniques that were applied, the audit result, and, when necessary, recommendations and best practices. When the processes are not in compliance, the report will include: the audit objective, the Artificial Intelligence techniques that were applied, the audit result, corrective actions, recommendations, and best practices. The IAP-R16 report will contain the implementation schedule with corrective actions, recommendations, and best practices, and suggested implementation dates.

[0271] Role of the AI ​​Team: The Ethics Committee and the Risk Team are responsible for approving the Reports.

[0272] Logical Flow Detail Figure 19: The IAP-M5 process will obtain information from the IAP-F, IAP-n, IAP-P, IAP-C, IAP-L, IAP-DH, and IAP-M processes (880). Through AI Inference, the audit will be performed (881). If the processes are in compliance and it is the first audit, the IAP-R15 report will be generated with the audit results. The first IAP-R16 report with the implementation schedule will also be generated. If the processes from the previous step are in compliance and it is not the first audit, the IAP-R15 and IAP-R16 reports will be updated with the new results, always maintaining information on actions not yet completed from previous audits. Both reports will be sent for approval by the Ethics Committee and Risk Team (883).

[0273] If the processes are not in compliance and it is the first audit, the IAP-R15 report will be generated with the audit results, including the corrections that should be made, recommendations, and best practices. The IAP-R16 report with an implementation schedule will also be generated. If the processes are not in compliance and it is not the first audit, the IAP-R15 and IAP-R16 reports will be updated with the new results, always maintaining information on actions not yet completed from previous audits. Both reports will be sent for approval by the Ethics Committee and Risk Team (884).

[0274] If the reports are approved, the information from the reports will be made available to the IAP-R process (886). Monitoring and Control will be Continuous (888).

[0275] If the reports are not approved, the teams must, through AI Inference, adjust, update, and approve the reports (887). Then the information from the reports will be made available to the IAP-R process (886). Monitoring and Control will be Continuous (888).

[0276] IAP-M6: Preventive Artificial Intelligence - Monitoring and Control Algorithmic Management

[0277] Overview: This process aims to implement Algorithmic Management. Algorithmic Management will perform algorithmic performance analysis through metrics. Technologies evolve, and new techniques will be developed and discovered. The metrics and techniques that can be used in IAP processes are not limited to those presented in this disclosure. Other metrics and techniques may be chosen and used according to the project developed and / or its evolution.

[0278] When it is not possible to detail the source code due to industrial and commercial secrecy, it will be necessary to explain the decisions made by the algorithms.

[0279] Prerequisites: Developed source code.

[0280] Reports: This process generates the IAP-R17 and IAP-R18 reports.

[0281] The IAP-R17 Report will present the results of the algorithmic performance analysis. The IAP-R18 Report will present the implementation schedule for corrective actions and list recommendations and best practices (if any).

[0282] Role of the AI ​​Team: The Ethics Committee and the Risk team will be responsible for approving the Reports.

[0283] Logical flow detail figure 20: The process is initiated by defining the performance metric (900).

[0284] Using AI Inference, algorithmic analysis (901) will be performed. This analysis will verify if any adjustments or corrections are needed to improve algorithmic performance. These corrections will not be made automatically. The suggested corrections will be included in the IAP-17 reports.

[0285] The IAP-R17 and IAP-R18 reports will be generated. These reports will be sent to the Ethics Committee and Risk Team for analysis and approval (902). If the report is approved, the report information will be sent to the IAP-R process (905). If the report is not approved, the teams: Ethics Committee and Risk Team must re-analyze, adjust, update, and approve the IAP-R17 and IAP-R18 reports (904). Then, the report information will be made available to the IAP-R process (905).

[0286] Monitoring and Control will be Continuous and carried out whenever requested (906).

[0287] IAP-M7 - Preventive Artificial Intelligence - Monitoring and Controlling the reduction of crime after the application of preventive methods.

[0288] Overview: The IAP-M7 process will monitor whether preventive methods are effective in preventing crime. It will show the results of the IAPs process and we will analyze the effectiveness of the processes in achieving the desired goal, which is to reduce the number of crimes. Furthermore, we will examine if there are any procedural deficiencies that need to be corrected. We will be able to analyze whether we should further enhance preventive methods to achieve the desired effectiveness. This process will be important for the IAP-M5 Auditability process.

[0289] If there is still no new data to perform the comparative analysis, this information should be included in the IAP-R19 report. When might this occur? It can occur when the prediction period has not yet started.

[0290] Prerequisites: The IAP-F1 and IAP-n processes must have been completed.

[0291] Reports: This process generates the IAP-R19 report, which contains comparative figures over months / years regarding increases and / or decreases in crime. When there is no new data to perform the comparative analysis, this information should be included in the report.

[0292] Role of the AI ​​Team: The Ethics Committee and the Risk team will be responsible for approving the Report.

[0293] Logical flow detail figure 21: The IAP-M7 process will obtain the information from the IAP-F1, IAP-n (920) processes.

[0294] It will be checked whether there is new data for comparative analysis (925).

[0295] If new data is available, using AI Inference, a comparative analysis will be performed between the initial data obtained at the beginning of the IAP processing and new data on the crime in question (921). In this way, we can analyze whether there has been an increase or decrease in crime. The IAP-R19 report will be generated and sent for approval by the Ethics Committee and Risk Team (922). The information from the report will be made available for the IAP-R process (923). Monitoring the increase or decrease in crime is an ongoing action (924).

[0296] If there is no new data, the IAP-R19 report will be generated and sent to the Ethics Committee and Risk Team (922). The information from the report will be made available for the IAP-R process (923). Monitoring the increase or decrease in crime is an ongoing action of the IAP processes (924).

[0297] IAP-R: Preventive Artificial Intelligence - Report Management

[0298] Overview: Its purpose is to Obtain / Provide / Create Reports.

[0299] This process is of fundamental importance because it allows teams access to process results, maintaining transparency and explainability.

[0300] To meet the needs of crime analysis, and if necessary for greater transparency, the format and content of the reports may be altered, for example: adding new columns and more detailed content. New reports may also be created if necessary for transparency and explanation of the crime analyzed.

[0301] Prerequisites: There are no prerequisites.

[0302] AI Team Performance: There is no performance from the teams.

[0303] Reports:

[0304] IAP-R1: Report generated by the IAP-F1 process. It indicates whether or not discriminatory biases were present, and which biases were blocked by the border control. This report will list: Columns analyzed by the database, technique applied, and result obtained, thus maintaining transparency and explainability.

[0305] IAP-R1A: Report generated by the IAP-F1 process. Lists the data rejected by the Border Authority.

[0306] IAP-R2: Report generated by the IAP-n process. This report lists crimes by criminal category with preventive methods that will be analyzed for enhancement and creation of new methods.

[0307] IAP-R3: Report generated by the IAP-P process. Lists the preventive methods that need to be strengthened.

[0308] IAP-R4: Report generated by the IAP-P1 process. Lists the preventive methods that were created.

[0309] IAP-R5: Report generated by the IAP-F2 process. It indicates whether or not discriminatory biases were present, and which biases were blocked by the border control. This report will list the variables used to analyze whether or not biases existed, thus maintaining transparency and explainability.

[0310] IAP-R6: Report generated by the IAP-L process. Lists information from the databases, i.e., legislation, legal framework, treaties, standards, public documents, doctrines, scientific research, and other legislation as defined. The list includes the name and date of the database update.

[0311] IAP-R7: Report generated by the IAP-DH process. Lists information from the databases, specifically Human Rights legislation. The list includes the name and date of the last update of the Legal Framework.

[0312] IAP-R8: Report generated by the IAP-C process. This report indicates whether the Preventive Methods violate the Laws.

[0313] IAP-R9: Report generated by the IAP-C process. This report contains the technical opinion of the IA Reviewer.

[0314] IAP-R10: Report generated by the IAP-C1 process. This report contains the Risk plan.

[0315] IAP-R11: Report generated by the IAP-C2 process. This report contains the Integrated Risk Plan, that is, the plan with risks identified by the IAP-C1 process and risks identified by the IAP-C2 process.

[0316] IAP-R12: Report generated by process IAP-M2. This report contains information on the crime classified after the legislation update.

[0317] IAP-R13: Report generated by process IAP-M2. This report contains information on the unclassified crime following the update of the legislation.

[0318] IAP-R14: Report generated by the IAP-M3 process. This report contains information on whether preventive methods are valid after the legislation update.

[0319] IAP-R15: Report generated by the IAP-M5 process. This report contains the audit results, including corrections, recommendations, and best practices.

[0320] IAP-R16: Report generated by the IAP-M5 process. This report contains the implementation schedule for the actions generated by the Audit.

[0321] IAP-R17: Report generated by the IAP-M6 process. This report contains the algorithmic performance results, including corrective actions that need to be implemented, recommendations, and best practices.

[0322] IAP-R18: Report generated by the IAP-M6 process. This report contains the implementation schedule for the actions generated by the algorithmic management.

[0323] IAP-R19: Report generated by process IAP-M7. This report contains information on increases or decreases in crime.

[0324] Logical Flow Detail Figure 22: The IAP-R process will obtain the process reports or store the reports in the Reports Database (1000). Depending on the crime analyzed, there may be a need to generate a new report to maintain transparency and explainability. If a new report is requested (1001), it will be generated through AI Inference and sent for approval by the Ethics Committee and Risk Team (1002). If the report is approved, the new report will be stored (1005). If the report is not approved, it will be necessary for the Ethics and Risk Committee to make adjustments through AI Inference (1004). Then the report will be stored (1005).

[0325] Examples of embodiments of the invention: two examples of embodiment will be shown. The first example pertains to the IAP Green Line Criminal Crimes Division, that is, crimes committed against the Environment. The second example pertains to the IAP Blue Line Criminal Crimes Division, that is, crimes committed against Public Faith.

[0326] Case Study 1: IAP-1 - WILDLIFE TRAFFICKING PREVENTION - Which crime will increase during the period of 1 º From January 2026 to December 31, 2027: Bird Trafficking or Primate Trafficking?

[0327] Overview: The illegal wildlife trade is already the 3rd ºThe largest illegal trade, ranked higher than drug trafficking and arms trafficking, is wildlife trafficking. This crime is a national and international concern. Cooperation between environmental protection organizations has been fundamental in helping to combat this monstrous trade. Defaunation leads to the extinction of species, affecting the entire ecosystem. The LAPs process will be applied to prevent wildlife trafficking.

[0328] Detailed description of IAP-1 processes

[0329] AI Inference: The techniques used were chosen based on the historical database for analysis; however, it is important to emphasize again that other techniques may be used depending on the characteristics and complexity of the historical database. In other words, the techniques involving Artificial Intelligence are not limited to those presented in this specific case. As already highlighted, the techniques will evolve, and the lAPs process will easily incorporate this evolution.

[0330] IAP-E:

[0331] Information is obtained from Databases (1100). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, a knowledge base was generated (1101).

[0332] IAP-I:

[0333] Knowledge base from IAP-E (1200) was obtained. Then information was obtained from the Databases (1201). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, knowledge base (1202) was generated.

[0334] IAP-A:

[0335] Knowledge base from IAP-E (1300) was obtained. Then knowledge base from IAP-I (1301) was obtained. Then information from the Databases (1302) was obtained. Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, knowledge base (1303) was generated.

[0336] IAP-F1:

[0337] The data (seized_specimen, seizure_term and seized_good) were obtained from the website https: / / dadosabertos.ibama.gov.br / dataset / (200). As per Item 4 º § 4 º of Article 4 ºUnder the General Data Protection Law (Law 13.709 of 2018), prior authorization for the use of collected data was not required. Three databases will be used: seized_property, seized_specimen, and seizure_term. The files are sent to the next stage.

[0338] Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, a critical analysis was performed on the data based on the discriminatory biases database (201).

[0339] The Discriminatory Biases database contains sensitive general data, such as age, race, and others. Discriminatory biases were analyzed using the Fairness-Aware Machine Learning technique. The metric used was the Disparate Impact Ratio (DIR).

[0340] After applying the technique, the resulting DIR value was <8, indicating possible discriminatory bias. Personal information was identified in the database, such as NOM_PESSOA_APREENSAO (Person's Name), CPF_CNPJ_PESSOA_APREENSAO (Person's CPF or CNPJ), and NUM_PROCESSO (202). The next step was then initiated. The data NOM_PESSOA_APREENSAO, CPF_CNPJ_PESSOA_APREENSAO, and NUM_PROCESSO were anonymized using the suppression technique and removed (207). The IAP-R1 report (shown in Figure 23) was generated and sent for approval by the Ethics Committee (208). Using Natural Language Processing, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee. The metrics used were Accuracy, Precision, and Confusion Matrix. After this analysis the report was approved (209), the border was opened and the data were made available for the IAP-R and IAP-n processes (205). The IAP-1 process was initiated (300).

[0341] IAP-n => IAP-1 (n=1 and 1=Wildlife Trafficking): IAP-1 - Green Criminal Line - Crimes committed against the environment.

[0342] The data are obtained from the IAP-F1 process (300). The defined prediction period was from 2026 to 2027 (301).

[0343] The next stage of exploratory data analysis (302) was then initiated. In this stage, bar and column graphs were plotted showing which animals were found dead and alive. Principal Component Analysis (PCA) was also performed. The PCA technique is able to identify trends, patterns, or outliers. The next stage (303) was then initiated.

[0344] Using AI Inference, it was analyzed which animals will be most sought after by traffickers. 70% of the data was reserved for training and 30% for testing. Time Series Analysis was performed with its respective decomposition of Residual, Seasonality, Trend, and Observed. The series is stationary. The Auto-ARIMA Machine Learning Model was applied with RMSE (Root Mean Square Error) and MAE (Mean Absolute Error) metrics. The XGBoost technique was applied, identifying that the applied model is a good model. It was identified that there will be an increase in crime in the search for birds (303). Subsequently, the IAP-R2 Report (shown in Figure 24) was generated for approval by the AI ​​Ethics Committee (304). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee. The metrics used were Accuracy, Precision, and Confusion Matrix.After this analysis the report was approved (305) and then the information was made available for the IAP-P, IAP-P1 and IAP-R processes (308). The IAP-P process was initiated (600).

[0345] IAP-P:

[0346] The data is obtained from process IAP-1 (600).

[0347] Using Natural Language Processing, Supervised Machine Learning with Deep Learning, the crime was linked to its respective preventive method from the Preventive Methods database (601). This resulted in the identification of preventive methods that need to be enhanced (603). Subsequently, the IAP-R3 Report (shown in Figure 25) was generated for approval by the Ethics Committee (604). Using Natural Language Processing, Supervised Machine Learning with Deep Learning, the Ethics Committee conducted an analysis. After this analysis, the report was approved (605) and the information was then made available for the IAP-C and IAP-R processes (607). The IAP-C process was initiated (700).

[0348] IAP-L:

[0349] The data were obtained from the Databases (500). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, a knowledge base was generated (501). The IAP-R6 Report (502) was generated (shown in figure 28). Then the information was made available to the IAP-C and IAP-M1 processes (503).

[0350] IAP-DH:

[0351] The data were obtained from the Databases (400). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, a knowledge base was generated (401). The IAP-R7 Report was generated (402) (shown in figure 29). Then the information is made available to the IAP-C and IAP-M1 processes (403)

[0352] IAP-P1:

[0353] Information was obtained from processes IAP-1 and IAP-P (620). IAP-1 shows information on the increase in bird trafficking and the IAP-P project shows information on preventive methods that need to be strengthened.

[0354] Using Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, new preventive methods for crime prevention were created (621). Subsequently, the IAP-R4 Report (shown in Figure 26) was generated for approval by the Ethics Committee (622). Using Natural Language Processing and Supervised Machine Learning with Deep Learning, the Ethics Committee conducted an analysis. After this analysis, the report was not approved (623) and sent for re-analysis by the Ethics Committee (604). The report was rejected because the new preventive method violated National Sovereignty. The Ethics Committee re-analyzed and excluded the rejected methods, thus approving the preventive methods.The information was then made available for the IAP-C and IAP-R processes (625).

[0355] IAP-F2:

[0356] The information was obtained from IAP-P1 (250). The objective was to analyze whether the new preventive methods have discriminatory biases. Using Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, a critical analysis of the data was performed based on the discriminatory biases database (251). After applying the technique, the resulting DIR value was =1, indicating that there is no discriminatory bias. The IAP-R5 report (shown in Figure 27) was generated and sent for approval by the Ethics Committee (253). Using Natural Language Processing, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee. The metrics used were Accuracy, Precision, and Confusion Matrix. After this analysis, the report was approved (254).The second frontier was opened and information was made available for the IAP-C and IAP-R processes (255).

[0357] IAP-C:

[0358] Information was obtained from processes IAP-L, IAP-DH, IAP-P, IAP-P1, and IAP-F2. This way, the Compliance process will have access to the knowledge bases generated on legislation, preventive methods, and the opening of the second frontier (700). Through Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, it was verified whether there was a violation of the laws. In this case, there was no violation of the laws. The IAP-R8 report (shown in Figure 30) was generated and sent to the Ethics Committee for approval (704). Through Natural Language Processing, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee. The metrics used were Accuracy, Precision, and Confusion Matrix. After this analysis the report was approved (705).The IAP-R8 report was sent to the AI ​​Reviewer so that he could prepare the IAP-R9 Report (shown in figure 31) (708). Using Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, the Report was approved by the reviewer (709) and the information was made available for the IAP-C1, IAP-C2 and IAP-R processes (711).

[0359] IAP-C1:

[0360] Information was obtained from the IAP-C process (720). Risk Analysis (721) was performed using Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks. To perform AI Inference, the Risk Database was accessed. This database contains information on probability, impact, final risk assessment, as well as the risk plan. In this crime, probability, impact, and final risk assessment were classified as Low, Medium, or High. Low probability was assigned to the unlikely risk of occurrence, thus receiving the value 1. Medium probability was assigned to the likely risk of occurrence, thus receiving the value 2. High probability was assigned to the almost certain risk of occurrence, thus receiving the value 3.Low impact was assigned to the risk that will only impact one IAP process, thus receiving a value of 1. Medium impact was assigned to the risk that impacts several IAP processes, but does not violate laws or human rights, thus receiving a value of 2. High impact was assigned to the risk whose impact is catastrophic with violation of laws and human rights, thus receiving a value of 3. The IAP-R10 report (shown in Figure 32) with the risk plan was generated and sent for approval by the Risk team (722). Using Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, an analysis was performed by the Risk Team. The metrics used were Accuracy, Precision, and Confusion Matrix. After this analysis, the report was approved, initiating risk treatment (726).Through Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, risk treatment was performed (726). Subsequently, the IAP-R10 report was updated and submitted for approval by the Ethics Committee (727). Through Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee. After this analysis, the report was approved (728). Subsequently, the information was made available for the IAP-M4, IAP-R, and IAP-C2 processes (730).

[0361] IAP-C2:

[0362] Information was obtained from the IAP-C and IAP-C1 processes (740). Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks were applied to identify new risks (741). The IAP-R11 report (shown in Figure 33) was generated with the updated Risk plan containing new risks identified by generative Artificial Intelligence. The report was sent for approval by the Risk team (742). This approval analysis was performed using Natural Language Processing and Supervised Machine Learning with Deep Learning. The metrics used were Accuracy, Precision, and Confusion Matrix. After this analysis, the report was approved (743). Subsequently, it was verified that a new risk had been identified (746), initiating risk treatment.Through Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, risk treatment was performed (747). Subsequently, the IAP-R11 report was updated and submitted for approval by the Ethics Committee (748). Through Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee. After this analysis, the report was approved (749) and the information was made available for the IAP-M4 and IAP-R processes (751).

[0363] IAP-M1:

[0364] Information was obtained from the IAP-L and IAP-DH processes (800). Legislation is updated, and therefore monitoring is necessary, such as, for example, Normative Instruction N º5 of May 13, 2021, which updated the Ibama Normative Instruction, N º 23 of 31 December 2014. Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, a knowledge base was generated (801). Then the information was made available to the Ethics team (802) and to the IAP-M2, IAP-M3 and IAP-M4 processes (803).

[0365] IAP-M2:

[0366] Information was obtained from process IAP-M1 (820). Legislation is updated, and therefore monitoring is necessary to check if the Crime of Animal Trafficking continues to be classified as a crime. Using Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, it was verified whether the crime continued to be classified as a crime (821). After verification, it was found that the crime continues to be classified as a crime (822). The IAP-R12 report was generated (shown in figure 34) (824). Subsequently, the information about the classified crime is made available to processes IAP-M3 and IAP-R (825).

[0367] IAP-M3:

[0368] Information was obtained from process IAP-M1 (840). Legislation is updated, and therefore monitoring is necessary to check if preventive methods remain valid and do not violate laws and human rights. In this case, a decree could arise prohibiting increased surveillance on trafficking routes so as not to endanger people living near these routes. The database of preventive methods contains information on the preventive methods identified by processes IAP-P and IAP-P1. Through Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks (841), it was verified whether the preventive methods remain valid (842).The IAP-R14 report (shown in figure 35) was generated to inform the Ethics Committee and Risk Team of the result (843). The information was then made available for the IAP-M4 and IAP-R processes (845).

[0369] IAP-M4:

[0370] Information was obtained from processes IAP-C1, IAP-C2, and IAP-M3 (860). In this way, information is obtained on the risks related to preventive methods to prevent wildlife trafficking. Through Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, it was verified whether the risk plan needed to be updated (861). After this analysis, it was found that the risk plans (IAP-R10 and IAP-R11) did not need to be updated (862). The names of those responsible and the start date were not yet defined. Continuous Monitoring should be carried out (868).

[0371] IAP-M5:

[0372] Information was obtained from processes IAP-F, IAP-1, IAP-P, IAP-C, IAP-L, IAP-DH and IAP-M to perform an audit (880). The audit was performed using Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks (881).

[0373] The IAP-F process documentation includes what the process is, its objective, the process steps, references to the IAP-R1, IAP-R1A, and IAP-R5 reports, information on the performance of the teams involved, the Artificial Intelligence techniques used, the name of the source database used, the names of the databases accessed, and the results obtained. It was verified that the reports were adequately completed. If any information was missing, it would imply non-compliance and the process would not pass the audit.

[0374] The IAP-1 process documentation includes what the process is, its objective, the process stages, the IAP-R2 report, the actions of the teams involved, the Artificial Intelligence techniques used, and the prediction period, which in the specific case of Wildlife Trafficking will be from January 2026 to December 2027 (inclusive). It was verified that the report was adequately completed. If any information were missing, it would imply non-compliance and the process would not pass the audit.

[0375] The IAP-P process documentation includes what the process is, its objective, the process steps, references to the IAP-R3 and IAP-R4 reports, information about the performance of the teams involved, the Artificial Intelligence techniques used, the name of the source database, and the names of the databases accessed. It was verified that the reports were adequately completed. If any information was missing, it would imply non-compliance and the process would not pass the audit.

[0376] The IAP-C process documentation includes what the process is, its objective, the process steps, references to the IAP-R8, IAP-R9, IAP-R10, and IAP-R11 reports, information on the performance of the teams involved, the Artificial Intelligence techniques used, and the results obtained. It was verified that the reports were adequately completed. If any information was missing, it would imply non-compliance, and the process would not pass the audit.

[0377] The IAP-L process documentation includes what the process is, its objective, the process steps, a reference to the IAP-R6 report, the Artificial Intelligence techniques used, the names of the databases accessed, and the databases used, which were described in the IAP-R6 report. It was verified that the report was adequately completed. If any information was missing, it would imply non-compliance, and the process would not pass the audit.

[0378] The IAP-DH process documentation includes what the process is, its objective, the process steps, a reference to the IAP-R7 report, the Artificial Intelligence techniques used, the names of the databases accessed, and the databases used, which were described in the IAP-R7 report. It was verified that the report was adequately completed. If any information was missing, it would imply non-compliance, and the process would not pass the audit.

[0379] The IAP-M process documentation includes what the process is, its objective, the process steps, and references to reports IAP-R12, IAP-R13, IAP-R14, IAP-R17, IAP-R18, and IAP-R19. It was verified that the reports were properly completed. If any information was missing, it would imply non-compliance and the process would not pass the audit.

[0380] After analysis by Artificial Intelligence, it was verified that the processes are in compliance (882). Subsequently, reports IAP-R15 and IAP-R16 were generated (shown in figures 36 and 37). The reports were sent for approval by the Ethics Committee and Risk Team (883), who, through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, analyzed and approved the reports (885). In the audit performed, all processes were approved. The information from the reports is made available to IAP-R (886). Perform new monitoring and control in 10 days (888).

[0381] IAP-M6:

[0382] The algorithmic performance metric (900) was defined. For performance analysis, asymptotic analysis was performed using Big O Notation Metrics, Code Profiling, Accuracy, Precision, Recall, and F1 Score. Asymptotic analysis is a technique used to analyze algorithmic behavior as the data size increases. The code profiling technique measures the execution time of different parts of the program. Precision, Recall, and F1 Score are techniques for measuring machine learning algorithms. Through Natural Language Processing and Machine Learning techniques, an algorithmic review was performed to analyze performance (901). We can observe that the execution time for binary search, which resulted in 0.000011 seconds, is much more efficient than linear search, which resulted in 0.010836 seconds. The Tottime values ​​in this case did not need to be optimized.The Precision, Recall, and F1-Score values ​​were above 80, indicating high precision, suggesting that the classifier's positive predictions are correct, and the F1-Score indicated a good balance. After algorithmic analysis, it was found that no corrections were necessary, as the performance result was good. Reports IAP-R17 and IAP-R18 were generated (shown in Figures 38 and 39). The reports were sent for approval to the Ethics Committee and Risk Team (902), which, using Natural Language Processing Technique and Supervised Machine Learning, analyzed and approved the reports. The information from the reports was made available to IAP-R (905). Continuous Monitoring is performed (906).

[0383] IAP-M7:

[0384] Information was obtained from processes IAP-F1 and IAP-1 (920). It was then checked whether there was new data for comparative analysis (925). In this specific case, preventive actions were not initiated and there is no new data for comparative analysis. Data from 2026 to 2027 have not yet been generated, as they are future data. Next, the IAP-R19 report (shown in Figure 40) was generated and sent to inform the Ethics Committee and Risk Team (922). Then, the information was made available for process IAP-R (923). Continuous Monitoring is carried out (924).

[0385] IAP-R:

[0386] During the execution of the lAPs processes, the reports were generated and managed by the IAP-R process. The reports were stored by IAP-R (1000). In this case, a new report was not requested (1001).

[0387] It is important to note that the content of the reports is being partially shown, but without hindering the understanding of this disclosure. TABLE 1 ART INTELLIGENCE: PREVENTIVE PHYSICAL - IAP-R1 REPORT [AP- Fl - Border 1 Data Source: https: / / dadosabertos.ibama.gov.br / dataset Technique Used - Discriminatory Biases Fairness-Aware Machine Learning Technique Metric: Disparate Impact Ratio [DLR] Result <-8. Discriminatory bias identified Result Details: Fields with personal information: NOM_PESSOA_APREENSAO, CPF_CNPJ_PERSON_SEIZED, PROCESS_NUMBER Technique Used for Correcting Anomalization with Suppression Technique employed by the Ethics Committee: Natural Language Processing, Supervised Machine Learning. Metric: Accuracy, Precision, and Confusion Matrix Final Result Open Border

[0388] This report presents the data source, the artificial intelligence technique used for discriminatory bias analysis, the result, artificial intelligence techniques used for correction against discriminatory biases, the artificial intelligence technique used by the Ethics Committee, and the final result. TABLE 2 PREVENTIVE ARTIFICIAL INTELLIGENCE - IAP-R2 REPORT IAP-1 – Crime Prediction by Criminal Line with Preventive Methods Previous process: IAP-F1 Forecast Period: January 1, 2026 to December 31, 2027 (inclusive) Historical reference base of the 2024 data. Exploratory Analysis: Bar and Column Charts Principal Component Analysis (PCA) Percentage of Training and Test Data: 70% Test: 30% Inference AI Time Series Analysis Auto-Armma XGBoost (eXtreme Gradient Boosting) Metrics: RMSE (Root Mean Square Error) and MAE (Mean Absolute Error), Final Result: Increase in Bird Trafficking

[0389] This report lists the previous process executed, the prediction period for the case in question, which graphs were used in the exploratory analysis, the percentage of data used for training and testing, which artificial intelligence techniques were used, and the final result, which will be an increase in bird trafficking. In this case, existing preventive methods have been listed and will also be verified in the next stage. TABLE 3 Preventive Artificial Intelligence - IAP-R3 Report IAP-P - Report with Preventive Methods to Enhance Previous Process: IAP-n (IAP-1) Inference AI Natural Language Processing Supervised Machine Learning Final Result Preventive Methods Environmental Education for Students Environmental education for teachers; Environmental education in communities living near protected areas; Awareness about reporting environmental issues. Extensive media coverage Social media engagement

[0390] Report on preventive methods that could be enhanced. This report lists the previous process executed, the artificial intelligence techniques used to identify which preventive methods need to be enhanced, and the final result. TABLE 4 Preventive Artificial Intelligence - IAP-R4 Report IAP-P1 - Report with Preventive Methods Created Previous process: IAP-n (IAP-1) and IAP-P Inference, AI, Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks — Transformers — Auto-Regressive Models — Recurrent Neural Networks. Preventive methods created Preliminary Result: Failed Preventive Methods Created - Post Reanalysis Final Result Approved

[0391] Report on preventive methods created by generative Artificial Intelligence. This report lists the preventive methods created and whether they were approved for not violating the country's legislation. This report presents the previous process executed, the artificial intelligence techniques that were employed, lists the preventive methods that were created, preliminary results, and final results after reanalysis. TABLE 5 Preventive Artificial Intelligence - IAP-R5 Report IAP-F2 - Border 2 Previous process: IAP-P1 Technique Used - Discriminatory Biases Fairness-Aware Machine Learning Technique Metric: Disparate Impact Ratio (DLR) Result: No discriminatory bias identified Result Details: No discriminatory bias identified Correction Technique Used: No correction Technique Employed by the Ethics Committee: Natural Language Processing. Supervised Machine Learning. Metrics: Accuracy, Precision, and Confusion Matrix. Final Result Open Border

[0392] Report on the results of the second frontier after the application of generative Artificial Intelligence. The frontier was opened indicating that there is no discriminatory bias. This report presents the previous process executed, the artificial intelligence techniques that were employed, the result, the details of the result, whether any techniques were employed to correct biases, techniques employed by the Ethics Committee, and the final result. https: / / jusbrasil.com.br / TABLE 6 Preventive Artificial Intelligence – IAP-R8 Report IAP-L - Report on Legislation Previous process IAP-P1 Reference for the database: https: / / ww4.planalto.gov.br / legislacao data https: / / www12.senado.leg.br / hpsenado https: / / www.trf3.jus.br https: / / www.lexml.gov.br / https: / / periodicos.capes.gov.br / htps: / ^3 jblsotec3-digital.stf._fos.br / xmluV https: / / catalogo.idp.edu.br / https: / / portal.stf.jus.br / https: / / cnj.jus.br https: / / www.trf1.jus.br / trf1 / home / https: / / www.trf2.jus.br, https: / / www.trf4.jus.br / ... Technique for Generating Natural Language Processing Supervised Machine Learning Base Knowledge Information Retrieval Systems Recommendation Systems Generative Neural Networks — Transformers – Auto-Regressive Models – Recurrent Neural Networks. Final Result: Knowledge Base Generated Legislation Database Law Description Constitution of the Federative Republic of Brazil Federal Constitution Brazilian legislation Ordering Brazilian Legal System: Law No. 9605 of 1998, Environmental Crimes Law Law No. 13.125 of 2015, Law on Access to Genetic Resources; Law No. 7.173 of 1983, Zoos Law Law No. 9613 of 1998 Money laundering Law No. 10,650 of 2003 Environmental Information Law Law No. 12,527 of 2011 Access to Information Law

[0393] Report on Legislation. This report presents the previous process executed, the reference of the databases used, the artificial intelligence techniques applied to generate the knowledge base, the final result with the generated knowledge base, and lists the legislation with their respective descriptions. TABLE 7 Preventive Artificial Intelligence – IAP-R7 Report IAP-DH - Report on Human Rights Legislation Reference for the database: https: / / www.un.org / data https: / / www.coe.int / en / web / portal / home https: / / www.amnesty.org / en / https: / / www.hrw.org https: / / www4.planalto.gov.br / legislacao / https: / / www12.senado.leg.br https: / / www.ov.br / nsdh / pt-br https: / / lexml.gov.br / https: / / jusbrasil.com.br / https: / / corteidh.or.cr / https: / / www.unicef.org https: / / ijrcenter.org / https: / / www.ohchr.org / en / ohchr_homepage... Technique for Generating Natural Language Processing Supervised Machine Learning Base Knowledge Information Retrieval Systems Recommendation Systems Generative Neural Networks — Transformers — Auto-Regressive Models — Recurrent Neural Networks. Final Result: Knowledge Base Generated Legislation: Legislation Approved by resolution AG / RES. 448 (IX-O / 79), adopted by the General Assembly of the OAS, at its Ninth Regular Session, held in La Paz, Bolivia, October 1979. Promulgates the Convention Against Torture and Decree No. 40, of 15 of Other Cruel Treatment or Punishment, February 1991 Inhuman or Degrading. Decree No. 7,037, of 21 Approves the National Rights Program December 2009 Human Rights Program - PNDH-3 and other provisions. Transforms the Council for the Defense of Human Rights into the National Council for Human Rights - CNDH; revokes Laws No. 4,319, of March 16, 1964, and No. 5,763, of December 15, 1971; and provides other provisions. June 2014. Measures to be taken. State Law No. 7,576, creating the State Council for the Defense of November 27th, Human Rights and other related provisions. 1991 related

[0394] Report on Human Rights Legislation. This report presents the references of the databases used, the artificial intelligence techniques that were used to generate the knowledge base, the final result with the generated knowledge base, and the list of legislation with their respective descriptions. TABLE 8 Preventive Artificial Intelligence – IAP-R8 Report IAP-C - Compliance Report Previous process; WL W4M IAP-R ÍAPd>Í and ÍAP-F2 Inference AI Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks - Transformers* - Auto-Regressive Models — Recurrent Neural Networks. Result Ps««ím:n>sr Approved Preventive Methods | Preventive Methods Environmental Education for Students b Environmental Education for Teachers Environmental education in communities living near protected areas. Raising awareness about reporting Extensive media coverage Social media engagement b Feasibility study for the development of animal feather sensors for use in airports b Creation of Public Policy to Encourage b Assisted Reproduction of Birds H Identify occurrences of psittacosis and xonoise in health units to investigate possible b centers of wildlife trafficking b Development of sensors based on b Spectroscope to be installed in areas along the trafficking route of wild auknais Final Result Approved

[0395] Report on process compliance. This report presents the previous processes that were executed, the artificial intelligence techniques that were employed to verify violations of laws, the preliminary result, the list of preventive methods, and the final result. TABLE 9 Preventive Artificial Intelligence - IAP-R9 Report iAP-C - Technical Opinion Previous process: IAP-L, IAP-DH, IAP-P, IAP-P1 and IAP-F2 Inference AI Natural Language Processing - Supervised Machine Learning - Information Retrieval Systems - Recommendation Systems - Generative Neural Networks - Transformers - Autoregressive Models - Recurrent Neural Networks. Introduction: The environment is one of the most important pillars for human life. This report aims to analyze, from a historical and constitutional legal perspective, the enhanced and developed preventive methods for preventing wildlife trafficking, specifically birds. Historical Analysis: The crime of animal trafficking is a phenomenon that has accompanied history, having emerged with the beginning of the exploitation of Brazilian lands. Currently, it generates... Legal-Constitutional Analysis: The 1988 Federal Constitution represents a major milestone for environmental protection. Article 225 of the Brazilian Constitution establishes that "Everyone has the right to an ecologically balanced environment, a common good of the people and essential to a healthy quality of life, imposing on the Public Authorities and the community the duty to defend and preserve it for present and future generations." Legislation has been created to establish criminal and administrative sanctions in the fight against wildlife trafficking. At the infraconstitutional level, there is no doubt that Law No. 9,605 of 1998 - the Environmental Crimes Law - was one of the greatest contributions on the subject. From a formal and substantive standpoint, I see no unconstitutionality in the matter under discussion... Vote Approved

[0396] Report containing the technical opinion generated by the expert trained by Artificial Intelligence. This report presents the previous process executed, the artificial intelligence techniques used by the expert, the introduction of the opinion, the historical analysis of the opinion, the Legal-Constitutional analysis, and the expert's vote. TABLE 10 Mim mvÈSWsS - SSlAfém Skis:: Is huh? ...ssSss.. ><:< 'X$:> VAW V _ s. ■s ««sios .-'W'iX ÍX? THE ?<..assisaafe;......................

[0397] Report on the risk analysis performed. This report presents the previous process executed, the artificial intelligence techniques that were applied, the creation of risks that may be Standard without application of Generative Artificial Intelligence or by Generative Artificial Intelligence, the risk description, the probability of the risk, the impact of the risk, the final risk assessment, the risk response, the action to be taken, the person responsible, the start date for addressing the risk, the end date and the status of the action. TABLE 11

[0398] Report on risk analysis generated by Generative Artificial Intelligence. This report presents the previous process executed, the artificial intelligence techniques that were applied, the creation of risks that may be Standard without application of Generative Artificial Intelligence or by Generative Artificial Intelligence, the risk description, the probability of the risk, the impact of the risk, the final risk assessment, the risk response, the action to be taken, the person responsible, the initial date to address the risk, the final date and the status of the action. TABLE 12 Preventive Artificial Intelligence - IAP-R12 Report IAP-M2 - Monitoring and Control Report on Crime Classification Previous process: IAP-M1 Inference, AI, Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks — Transformers - Auto-Regressive Models - Recurrent Neural Networks. Result: Classified Crime

[0399] Report on whether wildlife trafficking remains a crime. This report presents the previous process, the artificial intelligence techniques that were implemented, and the final result. TABLE 13 Preventive Artificial Intelligence - IAP-R14 Report IAP-M3 - Monitoring and Control Report on Preventive Methods Previous process: IAP-M1 Inference AI Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Result Valid Preventive Methods

[0400] Report on the outcome of preventive methods remaining valid. This report presents the previous process executed, the artificial intelligence techniques applied, and the result. TABLE 14 PREVENTIVE ARTIFICIAL INTELLIGENCE - IAP-R15 IAP-M5 REPORT - Monitoring and Control Report - Audit of Audited Processes IAP-F, IAP-1, IAP-P, IAP-Ç, IAP-L, IAP-DH and IAP-M. Audit Objective: To verify if all reports are up-to-date. Inference AI Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks — Transformers - Auto-Regressive Models - Recurrent Neural Networks. IAP-F Documentation Approved IAP-1 Documentation Approved IAP-P Documentation Approved IAP-C Documentation Approved IAP-L Documentation Approved IAP-DH Documentation Approved IAP-M Documentation Approved Audit Result Approved

[0401] Audit report. This report presents the list of audited processes, the audit objective, the artificial intelligence techniques employed, and the result regarding the approval of the audited documentation. TABLE 15 Preventive Artificial Intelligence - IAP-R16 IAP-M5 Report - Monitoring and Control Report - Implementation of Audited Processes IAP-F, IAP-1, IAP-P, IAP-C, IAP-L, IAP-DH and IAP-M. AI Inference. Natural Language Processing. Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks — Transformers - Auto-Regressive Models - Recurrent Neural Networks. Implementation Plan Not Applicable

[0402] Report on the implementation plan. This report presents the processes that were audited for reference, the artificial intelligence techniques used, and the result regarding whether there is a need to generate the implementation plan. TABLE 16 Preventive Artificial Intelligence - IAP-17 Report IAP-M6 - Monitoring and Control Report - Algorithmic Performance Programming Language Python Inference AI Natural Language Processing: Machine Learning Techniques for analyzing asymptotic analysis with metrics, Big 0 notations, profiling, code performance, accuracy, precision, recall, and FlScore. IAP-F1 Performance Analysis Big 0 Notation Linear Search: 0.010836 seconds Binary Search: 0.000011 seconds Nealis Code Profiling: 1 Total: Loss H Accumulation 0.409 0.409 0.409 seconds seconds seconds Accuracy 0.87 Precision 0.8596 Recall 0.8941 FL Score 0.8755 IAP-P Performance Analysis Big 0 Notation Linear Search: 0.010836 seconds Binary Search: 0.000014 seconds Ncalls Code Profiling: 1 Total Time: Percall Cumtime 0.420 0.420 0.420 seconds seconds seconds Accuracy 0.89 Precision 0.8696 Recall 0.8840 FlScore 0.8965

[0403] Report on algorithmic performance results. This report presents the programming language used, the artificial intelligence techniques applied, and the results of the performance analysis. TABLE 17 Preventive Artificial Intelligence - IAP-R18 Report IAP-M6 - Monitoring and Control Report — Corrective Actions for Algorithmic Performance Inference AI Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Implementation Plan - Audit Approved Corrective Actions No corrective actions Best Practices: Conduct a new audit in 10 days.

[0404] This report presents the Artificial Intelligence techniques employed, whether an implementation plan was created, the corrective actions (in this case, none), and the recommended best practices. TABLE 18 Preventive Artificial Intelligence - IAP-R19 Report IAP-M7 - Monitoring and Control Report - AI Inference Performance Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks — Transformers - Auto-Regressive Models - Recurrent Neural Networks. Comparative analysis period. Reference year: 2024 Year for Comparison: 2026 to 2027 Information regarding Reduction or No data for analysis Increase in Crime

[0405] Report on crime reduction. This report presents the Artificial Intelligence techniques that were employed, the period for comparative analysis, and information on crime reduction or increase.

[0406] Case Study 2: IAP-2 - PREVENTION OF BANKNOTE COUNTERFEITING - In which State will there be an increase in banknote counterfeiting between the period of 1 º From January 2026 to December 31, 2028?

[0407] Overview:

[0408] Counterfeiting is a transnational crime. Recent statistics show that more than 36% of the Brazilian population has received counterfeit banknotes. The Federal Police have been conducting operations to combat this crime. There was a reduction in counterfeiting between 2023 and 2024 due to Federal Police operations that dismantled counterfeiting workshops. However, even with all these efforts, this crime continues to occur and challenge Brazilian authorities.

[0409] AI Inference: The techniques used were chosen based on the historical database for analysis; however, it is important to emphasize that other techniques may be used depending on the characteristics and complexity of the historical database. In other words, the techniques involving Artificial Intelligence are not limited to those presented in this specific case.

[0410] Detailed description of IAP-2 processes

[0411] IAP-E:

[0412] Information is obtained from Databases (1100). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, a knowledge base was generated (1101).

[0413] IAP-1:

[0414] Knowledge base from IAP-E (1200) was obtained. Then information was obtained from the Databases (1201). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, knowledge base (1202) was generated.

[0415] IAP-A:

[0416] Knowledge base from IAP-E (1300) was obtained. Then knowledge base from IAP-I (1301) was obtained. Then information from the Databases (1302) was obtained. Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, knowledge base (1303) was generated.

[0417] IAP-F1

[0418] The data (FalsificacaoJDadosAbertos) were obtained from the website https: / / dadosabertos.bcb.gov.br / dataset / falsificacoes-por-ano-e-por-estado (200). As per Item 4 º § 4 º of Article 4 º Under the General Data Protection Law (Law 13.709 of 2018), prior authorization for the use of collected data was not required. The files are sent to the next stage.

[0419] The Discriminatory Biases database contains sensitive general data, such as age, race, among others. In this case, the database used for analysis did not contain data for comparability with the Discriminatory Biases database. The Pearson Correlation Coefficient technique was used. This technique verifies if there is a correlation between variables. The result of the Pearson correlation coefficient (r) was = 0, indicating that there is no relationship between the variables. Thus, it was considered that there are no discriminatory biases, proceeding to the next step (203).

[0420] The IAP-R1 report (shown in Figure 41) was generated and submitted for approval by the Ethics Committee. Using Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee (203). The metrics used were Accuracy, Precision, and Confusion Matrix. After this analysis, the report was approved (204), the frontier was opened, and the data were made available for the IAP-R and IAP-N processes (205). The IAP-2 process was initiated (205).

[0421] 2) IAP-n => IAP-2 (n=2 and 2=Counterfeiting of Banknotes): IAP-2 - Blue Criminal Line - Crimes Committed Against Public Faith.

[0422] The data are obtained from the IAP-F1 process (300). The prediction period defined was January 1, 2026 to December 31, 2028, that is, in which State will there be an increase in the counterfeiting of banknotes between January 1, 2026 and December 31, 2028? (301).

[0423] The next stage of exploratory data analysis (302) was then initiated. In this stage, bar and column graphs were plotted showing which states had the highest number of counterfeit banknotes. Principal Component Analysis (PCA) was also performed. The PCA technique is capable of identifying trends, patterns, or outliers. The next stage (303) was then initiated.

[0424] Using AI Inference, it was analyzed which states will experience an increase in counterfeit banknotes. 70% of the data was reserved for training and 30% for testing. Time Series Analysis was performed with its respective decomposition of Residual, Seasonality, Trend, and Observed. The series is stationary. The Auto-ARIMA Machine Learning Model was applied with RMSE (Root Mean Square Error) and MAE (Mean Absolute Error) metrics. The XGBoost technique was applied, identifying that the applied model is a good model. It was identified that there will be an increase in counterfeit currency in the State of São Paulo. (303). Subsequently, the IAP-R2 Report (shown in Figure 42) was generated for approval by the AI ​​Ethics Committee (304). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee (304). The metrics used were Accuracy, Precision, and Confusion Matrix.After this analysis the report was approved (305) and then the information was made available for the IAP-P, IAP-P1 and IAP-R processes (308). The IAP-P process was initiated (600).

[0425] 3) IAP-P:

[0426] The data is obtained from process IAP-1 (600).

[0427] Using Natural Language Processing, Supervised Machine Learning with Deep Learning, the crime was linked to its respective preventive method from the Preventive Methods database (601). This resulted in the identification of preventive methods that need to be strengthened (603). Subsequently, the IAP-R3 Report (shown in Figure 43) was generated for approval by the Ethics Committee (604). Using Natural Language Processing, Supervised Machine Learning with Deep Learning, the Ethics Committee conducted an analysis. After this analysis, the report was approved (605) and the information was then made available for the IAP-C and IAP-R processes (607). The IAP-C process was initiated (700).

[0428] 4) IAP-L:

[0429] The data were obtained from the Databases (500). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, a knowledge base was generated (501). The IAP-R6 Report (502) was generated (shown in figure 46). Then the information was made available to the IAP-C and IAP-M1 processes (503)

[0430] 5) IAP-DH:

[0431] The data were obtained from the Databases (400). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, a knowledge base was generated (401). The IAP-R7 Report was generated (402) (shown in figure 47). Then the information is made available to the IAP-C and IAP-M1 processes (403)

[0432] 6) IAP-P1:

[0433] Information was obtained from processes IAP-1 and IAP-P (620). IAP-1 shows information on the increase in the crime of counterfeiting banknotes in the State of São Paulo and the IAP-P project shows information on preventive methods that need to be strengthened.

[0434] Using Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, new preventive methods for crime prevention were created (621). The IAP-R4 Report (shown in Figure 44) was then generated for approval by the Ethics Committee (622). Using Natural Language Processing and Supervised Machine Learning with Deep Learning, the Ethics Committee conducted an analysis. After this analysis, the report was approved (623). The information was then made available for the IAP-C and IAP-R processes (625).

[0435] IAP-F2:

[0436] The information was obtained from IAP-P1 (250). The objective was to analyze whether the new preventive methods have discriminatory biases. The Pearson Correlation Coefficient technique was used. This technique verifies if there is a correlation between variables. The result of the Pearson correlation coefficient (r) was = 0, indicating that there is no relationship between the variables and therefore there are no discriminatory biases (252). The IAP-R5 report (shown in figure 45) was generated and sent for approval by the Ethics Committee (253). Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, an analysis was carried out by the Ethics Committee. The metrics used were Accuracy, Precision and Confusion Matrix. After this analysis, the report was approved (254). The second frontier was opened and the information was made available for the IAP-C and IAP-R processes (255).

[0437] IAP-C:

[0438] Information was obtained from processes IAP-L, IAP-DH, IAP-P, IAP-P1, and IAP-F2. This way, the Compliance process will have access to the knowledge bases generated on legislation, preventive methods, and the opening of the second frontier (700). Through Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, it was verified whether there was a violation of the laws. In this case, there was no violation of the laws. The IAP-R8 report (shown in Figure 48) was generated and sent to the Ethics Committee for approval (704). Through Natural Language Processing, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee. The metrics used were Accuracy, Precision, and Confusion Matrix. After this analysis the report was approved (705).The IAP-R8 report was sent to the AI ​​Reviewer so that he could prepare the IAP-R9 Report (described in figure 49) (708). Using Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, the Report was approved by the reviewer (709) and the information was made available for the IAP-C1, IAP-C2 and IAP-R processes (711).

[0439] IAP-C1:

[0440] Information was obtained from the IAP-C process (720). Risk Analysis (721) was performed using Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks. To perform AI Inference, the Risk Database was accessed. This database contains information on probability, impact, final risk assessment, as well as the risk plan. In this crime, probability, impact, and final risk assessment were classified as Low, Medium, or High. Low probability was assigned to the unlikely risk of occurrence, thus receiving the value 1. Medium probability was assigned to the likely risk of occurrence, thus receiving the value 2. High probability was assigned to the almost certain risk of occurrence, thus receiving the value 3.Low impact was assigned to risks that will only affect one IAP process, thus receiving a value of 1. Medium impact was assigned to risks that affect several IAP processes, but do not violate laws or human rights, thus receiving a value of 2. High impact was assigned to risks with catastrophic impacts that violate laws and human rights, thus receiving a value of 3. The IAP-R10 report (shown in Figure 50) with the risk plan was generated and sent for approval by the Risk team (722). Using Natural Language Processing, Supervised Machine Learning with Deep Learning techniques, the Risk Team performed an analysis. The metrics used were Accuracy, Precision, and Confusion Matrix. After this analysis, the report was approved, initiating risk treatment.Through Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, risk treatment was performed (726). Subsequently, the IAP-R10 report was updated and submitted for approval by the Ethics Committee (727). Through Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, an analysis was performed by the Ethics Committee. After this analysis, the report was approved (728). Subsequently, the information was made available for the IAP-M4, IAP-R, and IAP-C2 processes (730).

[0441] IAP-C2:

[0442] Information was obtained from processes IAP-C and IAP-C1 (740). Using Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, an analysis was performed to identify new risks (741). The IAP-R11 report (shown in Figure 51) was generated with information about the Risk plan. The report was sent for approval by the Risk team (742). Using Natural Language Processing, Supervised Machine Learning with Deep Learning, an analysis was performed by the Risk Team. The metrics used were Accuracy, Precision, and Confusion Matrix. After this analysis, the report was approved (743). No new risks were identified (746). Subsequently, the information was made available to processes IAP-M4 and IAP-R (751).

[0443] IAP-M1:

[0444] Information was obtained from the IAP-L and IAP-DH processes (800). The legislation is updated. Through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks, a knowledge base was generated (801). Then the information was made available to the Ethics team (802) and to the IAP-M2, IAP-M3 and IAP-M4 processes (803).

[0445] IAP-M2:

[0446] Information was obtained from process IAP-M1 (820). Legislation is updated, and therefore monitoring is necessary to check if the crime of counterfeiting banknotes continues to be classified as a crime. Using Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, it was verified whether the crime continued to be classified as a crime (821). After verification, it was found that the crime continues to be classified as a crime (822). The IAP-R12 report was generated (shown in figure 52) (824). Subsequently, the information about the classified crime is made available to processes IAP-M3 and IAP-R (825).

[0447] IAP-M3:

[0448] Information was obtained from process IAP-M1 (840). Legislation is updated, and therefore monitoring is necessary to check if preventive methods remain valid and do not violate laws and human rights. The database of preventive methods contains information on the preventive methods identified by processes IAP-P and IAP-P1. Through Natural Language Processing Techniques, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks (841), it was verified that the preventive methods remain valid (842). The IAP-R14 report (shown in Figure 53) was generated to inform the Ethics Committee and Risk Team of the result (843). The information was then made available for processes IAP-M4 and IAP-R (845).

[0449] IAP-M4:

[0450] Information was obtained from processes IAP-C1, IAP-C2, and IAP-M3 (860). This provided information on the risks related to preventive methods for preventing the crime of counterfeiting banknotes. Using Natural Language Processing, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation Systems, Generative Neural Networks, Transformers, Auto-Regressive Models, and Recurrent Neural Networks, it was verified whether the risk plan needed updating (861). After this analysis, it was found that the risk plans (IAP-R10 and IAP-R11) did not need updating (862). The names of those responsible and the start date were not yet defined. A new monitoring session was scheduled for one week later (868).

[0451] IAP-M5:

[0452] Information was obtained from processes IAP-F, IAP-1, IAP-P, IAP-C, IAP-L, IAP-DH and IAP-M to perform an audit (880). The audit was performed using Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, Information Retrieval Systems, Recommendation System, Generative Neural Networks, Transformers, Auto-Regressive Models, Recurrent Neural Networks (881).

[0453] The IAP-F process documentation includes what the process is, its objective, the process steps, references to the IAP-R1, IAP-R1A, and IAP-R5 reports, information on the performance of the teams involved, the Artificial Intelligence techniques used, the name of the source database used, the names of the databases accessed, and the results obtained. It was verified that the reports were adequately completed. If any information was missing, it would imply non-compliance and the process would not pass the audit.

[0454] The IAP-1 process documentation includes what the process is, its objective, the process stages, the IAP-R2 report, the actions of the teams involved, the Artificial Intelligence techniques used, and the prediction period, which in the specific case of the Counterfeit Banknote Crime will be from January 2026 to December 2028 (inclusive). It was verified that the report was adequately completed. If any information were missing, it would imply non-compliance and the process would not pass the audit.

[0455] The IAP-P process documentation includes what the process is, its objective, the process steps, references to the IAP-R3 and IAP-R4 reports, information about the performance of the teams involved, the Artificial Intelligence techniques used, the name of the source database, and the names of the databases accessed. It was verified that the reports were adequately completed. If any information was missing, it would imply non-compliance and the process would not pass the audit.

[0456] The IAP-C process documentation includes what the process is, its objective, the process steps, references to the IAP-R8, IAP-R9, IAP-R10, and IAP-R11 reports, information on the performance of the teams involved, the Artificial Intelligence techniques used, and the results obtained. It was verified that the reports were adequately completed. If any information was missing, it would imply non-compliance, and the process would not pass the audit.

[0457] The IAP-L process documentation includes what the process is, its objective, the process steps, a reference to the IAP-R6 report, the Artificial Intelligence techniques used, the names of the databases accessed, and the databases used, which were described in the IAP-R6 report. It was verified that the report was adequately completed. If any information was missing, it would imply non-compliance, and the process would not pass the audit.

[0458] The IAP-DH process documentation includes what the process is, its objective, the process steps, a reference to the IAP-R7 report, the Artificial Intelligence techniques used, the names of the databases accessed, and the databases used, which were described in the IAP-R7 report. It was verified that the report was adequately completed. If any information was missing, it would imply non-compliance, and the process would not pass the audit.

[0459] The IAP-M process documentation includes what the process is, its objective, the process steps, and references to reports IAP-R12, IAP-R13, IAP-R14, IAP-R17, IAP-R18, and IAP-R19. It was verified that the reports were properly completed. If any information was missing, it would imply non-compliance and the process would not pass the audit.

[0460] After analysis by Artificial Intelligence, it was verified that the processes are in compliance (882). Subsequently, reports IAP-R15 and IAP-R16 were generated (shown in figures 54 and 55). The reports were sent for approval by the Ethics Committee and Risk Team (883), who, through Natural Language Processing Technique, Supervised Machine Learning with Deep Learning, analyzed and approved the reports (885). In the audit performed, all processes were approved. The information from the reports is made available to IAP-R (886). Perform new monitoring and control in 7 days (888).

[0461] IAP-M6:

[0462] The algorithmic performance metric (900) was defined. For performance analysis, asymptotic analysis was performed using Big O Notation Metrics, Code Profiling, Accuracy, Precision, Recall, and F1 Score. Asymptotic analysis is a technique used to analyze algorithmic behavior as the data size increases. The code profiling technique measures the execution time of different parts of the program. Precision, Recall, and F1 Score are techniques for measuring machine learning algorithms. Through Natural Language Processing and Machine Learning techniques, an algorithmic review was performed to analyze performance (901). We can observe that the execution time for binary search, which resulted in 0.000011 seconds, is much more efficient than linear search, which resulted in 0.010836 seconds. The Tottime values ​​in this case did not need to be optimized.The Precision, Recall, and F1-Score values ​​were above 80, indicating high precision, suggesting that the classifier's positive predictions are correct, and the F1-Score indicated a good balance. After algorithmic analysis, it was found that no corrections were necessary, as the performance result was good. The IAP-R17 and IAP-R18 reports were generated (shown in Figures 56 and 57). The reports were sent for approval to the Ethics Committee and Risk Team (902), who, using Natural Language Processing Technique and Supervised Machine Learning, analyzed and approved the reports (903). The information from the reports was made available to IAP-R (905). Continuous Monitoring is performed (906).

[0463] IAP-M7:

[0464] Information was obtained from processes IAP-F1 and IAP-2 (920). It was then checked whether there was new data for comparative analysis (925). In this specific case, preventive actions were not initiated and there is no new data for comparative analysis. Data from 2026 to 2028 have not yet been generated, as they are future data. Next, the IAP-R19 report (shown in Figure 58) was generated and sent to inform the Ethics Committee and Risk Team (922). Then, the information was made available for process IAP-R (923). Continuous Monitoring is carried out (924).

[0465] IAP-R:

[0466] During the execution of the lAPs processes, reports are generated and managed by the IAP-R process. The reports were stored by IAP-R (1000). In this case, a new report was not requested (1001).

[0467] It is important to note that the content of the reports is being partially shown, but without hindering the understanding of this disclosure. TABLE 19 Preventive Artificial Intelligence - IAP-R1 Report SAP-Fl - Border 1 Data Source: Central Bank of Brazil https: / / dadosabertos.bcb.gov.br / dataset / falsificacoes-por-ano-e-por-estado Technique Employed — Biases Pearson Correlation Coefficient Discriminatory Result r=0 Result: There are no discriminatory biases. Result Details: There is no correlation between the variables. Technique used for correction: Not applicable. Technique employed by the Ethics Committee: Natural Language Processing, Supervised Machine Learning. Metric: Accuracy, Precision, and Confusion Matrix Final Result Open Border

[0468] This report presents the IAP-R1 report of the first frontier of the concrete case IAP-2 for the Prevention of the Crime of Counterfeiting Banknotes. This report presents the data source, the artificial intelligence technique used for discriminatory bias analysis, the result, artificial intelligence techniques used for correction against discriminatory biases, the artificial intelligence technique used by the Ethics Committee, and the final result. TABLE 20 g OG É S® OTCI «WWW ~ AStATÓPIO sAP-02 iA -1 - Crime Prediction by Criminal Line with Preventive Methods previous Process. iÁRFl PeHodo pao Freddie 1'? give j de 2G2S stê §1 give deaerntas de Ba&e de r*$er èn*:sa Rsádei-as dos dsdos 2G22 sté 2024 dfscksstve} iVsèisse BMpioratòoa: Grecos Bars and Columns PCA Analysis % of Gsdõs and Te^e Training Test. SOO; ieíferênde IA Anêilse Série Temporal AiAõ-ARrw XGSsosi («Xtrnw GrsdiefA Mêódox; kMSS tRoot Square Erw> and ] BUT (Mean Absolute £JW Rseufedo Rrseí Ar.sí!"«nUs d« fehíÁCAçéo de Cediul&s rso Estad o of Saint Per

[0469] Report on crime prediction. This report presents the previous process that was executed, the prediction period for the case in question, the base year of reference for the data being used for prediction, which graphs were used in the exploratory analysis, the percentage of data used for training and testing, which artificial intelligence techniques were used, and the final result. In this case, existing preventive methods were listed and will also be verified in the next stage. TABLE 21 Preventive Artificial Intelligence - IAP-R3 Report IAP-P - Reteiorto cans Preventive methods to Potentiate Previous processes: iAP-n OAP-1) Inference, AI, Natural Language Processing Supervised Machine Learning Result F sna! Preventive Methods: Audit of Cash Acquisition Contracts Internal audits in the bodies of Cash control monitoring of activities related to the destruction of cash. To raise awareness among citizens about the complaints investigations into cybercrimes...

[0470] This is the IAP-R3 report on which preventive methods can be enhanced from the specific case IAP-2 for the Prevention of Counterfeit Banknotes. This report lists the previous process executed, the artificial intelligence techniques used to identify which preventive methods will need to be enhanced, and the final result. TABLE 22 Preventive Artificial Intelligence - IAP-R4 Report IAP-P1 - Report with Preventive Methods Created Previous process: IAP-n (IAP-1) and IAP-P Inference, AI, Natural Language Processing, Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Preventive methods created Final Result Approved

[0471] This report lists the preventive methods created and approved for not violating the country's legislation. This report presents the previous process executed, the artificial intelligence techniques that were employed, lists the preventive methods that were created, and the final result. TABLE 23 Preventive Artificial Intelligence – IAP-R5 Report IAP-F2 - Border 2 Previous process: IAP-P1 Technique Employed – Discriminatory Biases Pearson Correlation Coefficient Result r=0 Result: There are no discriminatory biases. Result Details: There is no correlation between the variables. Technique used for correction: Not applicable. Technique employed by the Ethics Committee: Natural Language Processing, Supervised Machine Learning. Metrics; Accuracy, Precision, and Confusion Matrix Final Result Open Border

[0472] This report states whether or not discriminatory biases were detected after the application of Generative Artificial Intelligence and which biases were blocked by Fronteira. This report presents the previous process executed, the artificial intelligence techniques that were employed, the result, the details of the result, whether any technique was employed to correct the biases, techniques employed by the Ethics Committee, and the final result. TABLE 24 Preventive Artificial Intelligence – IAP-R6 Report IAP-L - Report on Legislation Previous process: iAP-Pi Reference for htt:p <s: / A«ww4 pianaft&.gsv:br / tegisiata0,. tiaclcs base https: / Aw»¥l 2.senado.ieg. br ht^s: / Aww.trf3jus.br / t http: / / ft , VA¥.iexrni.gov.br / ,. https: / / 3pea.p5v.br / portal / , https: / / www.jusbrasil.com.br / ht^2s: / Aww.periodicos.cape5.gov.br / , ^•stps^' / bitoUDSjecaeijgstal.-s-'tf.jus.ajr / KmliiiiZ, tttps; / / íat31cgo. sdp.edu. sr / ?_gi=l*lsb59ey*_gcs_3í.i*h'ITAwMzgwNjiy'Ni4K NzM3MzcwMDE4, htps: / / portaLstf.jus.br / , lttps: / / www..cnj.ju$.br / , htps: / Asww..si| jus.br / https:: / / www..trfI.jus.brArfl / horne / https: / Aww.trf2jiis.br / , https: / / mw.trf4.jus. br / .... Inferènoa I. Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks - Transformers - Auto-Regression Models - Recurrent Neural Networks. Final Result: Knowledge base generated Data Legislation Description Constitution of Federal Republic Federal Constitution of &rasii Legislation of Brazil, Regulation Decree-Law No. Brazilian Law 2.848 / 1940 Penal Code Organization Law Law No. 12.850 / 2013 Criminal Money Laundering Law Law No. 9.613 / 1998 Money Child Statute and Law No. 8.069 / 1990 on Adolescents Law on Crimes Against the Financial System National Law No. 7,492 / 1986 System Law Law No. 4,595 / 1964 National Finance Law of Interception Law No. 9.296 / 1996 Telefônica

[0473] This report presents the previous process executed, the reference databases used, the artificial intelligence techniques applied, the final result with the generated knowledge base, and lists the legislation with their respective descriptions. TABLE 25 Preventive Artificial Intelligence – IAP-R7 Report IAP-DH - Report on Human Rights Legislation Reference for database https: / / www..ü n. ar§ / htps: / / www;coeJnt / ep / web / 'portil / home https: / / www.amnesty.srg / en / https: / / www.hrw.org htps: / / www.4.pianaito. gpv.br / legislac3o / https: / / wtwl2.senadoJeg.br https: / / wT.wr.ov.br / mdh / pt-br https: / / iexmi.gov.br / https: / / jusbrasil.com.br / https: / / co-rteid h. or.cr / https: / / www.smicef.org https: / / lrcester.org / ht ps: / / w(w. ohchr. org / en / ohchr_homepage... Technique for Generating Natural Language Processing Supervised Machine Learning Sase Knowledge Information Retrieval Systems Recommendation Systems Generative Neural Networks - Transformers - Auto-Regressive Models — Recurrent Neural Networks. Final Result: Knowledge Base Generated Legislation: Legislation related to Human Rights Statute of the Inter-American Court of Human Rights Approved by resolution AG / RES. 448 (IX-O / 79), adopted by the General Assembly of the OAS, at its Ninth Regular Session, held in La Paz, Bolivia, October 1979. Promulgates the Convention Against Torture and Decree No. 40, of February 15, 1991, concerning Other Cruel, Inhuman or Degrading Treatment or Punishment. Decree No. 7,037, of December 21, 2009 Approves the National Human Rights Program - PNDH-3 and provides other measures. Transforms the Council for the Defense of Human Rights into the National Council for Human Rights - CNDH; revokes Laws No. 4,319, of March 16, 1964, and No. 5,763, of December 15, 1971; and provides other measures. State Law No. 7,576, of November 27th, creates the State Council for the Defense of Human Rights and provides other measures. 1991 related

[0474] This report presents the reference databases used, the artificial intelligence techniques employed to generate the knowledge base, the final result with the generated knowledge base, and the list of legislation with their respective descriptions (TABLE 26). Preventive Artificial Intelligence – IAP-R8 Report IAP-C - Compliance Report Previous process: IAP-L, IAP-DH, IAP-P, IAP-P1 and IAP-F2 Inference AI Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks – Transformers – Auto-Regressive Models – Recurrent Neural Networks. Preliminary Result Approved Preventive Methods CssríVaS©» àss ds áds sj -S <í:gôv The f***************<^\********X^^ ds* fk-síjipçs* reiác<on*fíí> $ -xs : '' > \ < S < íAÔS ; «>? Wis?s£s$ iáiSSádss WS 1 SSss f; esçs -ís sdWss 1 s wWsfe SráAsss ©í ■«" § Final Result Approved

[0475] This report presents the previous processes that were executed, the artificial intelligence techniques that were employed to verify violations of the laws, the preliminary result, the list of preventive methods, and the final result. TABLE 27 PREVENTIVE ARTIFICIAL SCIENCE - RELCTOPHILO IAP-R9 IAP-C - Technical Opinion Previous process: i rr> IAP-L, IAP-DH, iAP-P, IAP-P1 and IAP-F2 Inference AI Natural Language Processing - Supervised Machine Learning - Information Retrieval Systems - Recommendation Systems - Generative Neural Networks - Transformers - Autoregressive Models - Recurrent Neural Networks. Counterfeiting banknotes is a crime that causes great harm to the government and society. The Central Bank faces the challenge of defining strategic plans to guarantee the supply of currency to society. The MECIR, the Currency Circulation Department of the Central Bank of Brazil, is responsible, among other functions, for monitoring currency counterfeiting... Historical Analysis: The crime of counterfeiting money historically developed around the 19th century. The routes chosen by counterfeiters were routes that offered easy access to transportation of the time, such as... Legal and Constitutional Analysis: The crime of counterfeiting banknotes is defined in Article 289. To counterfeit, fabricate, or alter metallic or paper currency of legal tender in the country or abroad: Penalty - imprisonment from three to twelve years, and a fine.Also subject to the same penalties is anyone who, on their own behalf or on behalf of another, imports or exports, acquires, sells, exchanges, transfers, lends, keeps, or introduces counterfeit currency into circulation. The constituent elements of this crime are the conduct, the material object, and intent. Vote Approved

[0476] This report presents the previous process executed, the artificial intelligence techniques used by the reviewer, the introduction of the opinion, the historical analysis of the opinion, the legal-constitutional analysis, and the reviewer's vote. TABLE 28 IAP-R10- IAP-2

[0477] This report presents the previous process executed, the artificial intelligence techniques that were applied, the creation of risks that may be Standard without the application of Generative Artificial Intelligence or by Generative Artificial Intelligence, the risk description, the probability of the risk, the impact of the risk, the final risk assessment, the risk response, the action that will be taken, the person responsible, the initial date to address the risk, the final date and the status of the action. TABLE 29 Preventive Artificial Intelligence – IAP-R11 Report IAP-C2 - Risk Report – Generative AI Previous process: IAP-C and IAP-C1 Inference, AI, Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks – Transformers – Auto-Regressive Models – Recurrent Neural Networks. Result: No new risks were identified.

[0478] This report presents the previous process executed, the artificial intelligence techniques used, and the result. TABLE 30 Preventive Artificial Intelligence - IAP-R12 Report IAP-M2 - Monitoring and Control Report on Crime Classification Previous process: IAP-M1 Inference, AI, Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Result: Classified Crime

[0479] This report presents the previous process executed, the artificial intelligence techniques that were applied, and the final result. TABLE 31 Preventive Artificial Intelligence - IAP-R14 Report IAP-M3 - Monitoring and Control Report on Preventive Methods Previous process: IAP-M1 Inference, AI, Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks — Transformers - Auto-Regressive Models - Recurrent Neural Networks. Result Valid Preventive Methods

[0480] This report presents the previous process executed, the artificial intelligence techniques applied, and the result. TABLE 32 PREVENTIVE ARTIFICIAL INTELLIGENCE - IAP-R15 IAP-M5 REPORT - Monitoring and Control Report - Audit of Audited Processes IAP-F, IAP-1, IAP-P, IAP-C, IAP-L, IAP-DH and IAP-M. Audit Objective: To verify if all reports are up-to-date. Inference AI Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks, Approved IAAP-F Documentation IAP-1 Documentation Approved IAP-P Documentation Approved IAP-C Documentation Approved Documentation lAP-L Approved IAP-DH Documentation Approved IAP-M Documentation Approved Audit Result Approved

[0481] This report presents the list of audited processes, the audit objective, the artificial intelligence techniques employed, and the result regarding the approval of the audited documentation. TABLE 33 PREVENTIVE ARTIFICIAL INTELLIGENCE - IAP-R15 IAP-M5 REPORT - Monitoring and Control Report - Implementation Audited Processes 1 APT, I. AP-1, IAP-P, IAP-C, IAP-L, IAP-DH and IAP-M AI Inference Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks — Transformers - Auto-Regressive Models - Recurrent Neural Networks. Implementation Plan Not Applicable

[0482] This report presents the processes that were audited for reference, the artificial intelligence techniques used, and the result regarding whether there is a need to generate an implementation plan. TABLE 34 Preventive Artificial Intelligence - IAP-R17 IAP-M6 Report - Monitoring and Control Report - Algorithmic Performance Programming Language: Python Inference, AI, Natural Language Processing Machine Learning Techniques for analyzing asymptotic analysis with metrics, Big 0 notations, profiling, code performance, accuracy, precision, recall, and FI Score. IAP-F1 Performance Analysis Big O Notation Linear Search: 0.010836 seconds Binary Search: 0.000011 seconds Profiling of Code N falls Is: 1 Total: Percaíl Cumtíme 0.409 0.409 0.409 seconds seconds seconds Accuracy 0.87 Precision 0.8595 Recall 0.8941 FL Score 0.8765 iAP-P Performance Analysis Big 0 Notation Linear Search: 0.010836 seconds Binary Search: 0.000014 seconds Profiling Code N calls: 1 Total time: Perca lí Cumtime 0.420 0.420 0.420 seconds seconds seconds Accuracy 0.89 Precision 0.8696 Recall 0.8840 FI Score 9.8965

[0483] This report presents the programming language used, the artificial intelligence techniques applied, and the results of the performance analysis. TABLE 35 Preventive Artificial Intelligence - IAP-R18 Report I P-M6 - Monitoring and Control Report - Corrective Actions for Algorithmic Performance Inference, AI, Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation System Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Challenge Plan - Audit approved Corrective Actions No corrective actions Best Practices: Conduct a new audit in 10 days.

[0484] This report presents the Artificial Intelligence techniques employed, whether an implementation plan was created, the corrective actions (in this case, none), and the recommended best practices. TABLE 36 Preventive Artificial Intelligence - IAP-R19 Report IAP-M7 - Monitoring and Control Report - Performance Inference in Natural Language Processing Supervised Machine Learning Information Retrieval Systems Recommendation Systems Generative Neural Networks - Transformers - Auto-Regressive Models - Recurrent Neural Networks. Comparative analysis period. Reference year: 2023 and 2024 Year for Comparison: 2026 to 2028 Information on Reduction or No data for anafce Increase in Crime

[0485] This report presents the Artificial Intelligence techniques that were employed, the period for comparative analysis, and information on the reduction or increase in crime.

Claims

MODIFIED CLAIMS Received by the International Secretariat on May 26, 2026 (26.05.2026) 1. A process for enhancing and creating preventive methods against crime, characterized by the integrated action of multiple teams formed by artificial intelligence, in order to produce autonomous and refined decisions, comprising the following stages: • a process for data processing and exploration configured to perform the collection, cleaning, normalization, tokenization, indexing and exploration of historical and current data on methods for crime prevention (1403); • a process for preparing artificial intelligence models by consolidating different types of Artificial Intelligence (1404), configured to train and test models, including multimodal integration of natural language processing (1400), machine learning (1401) and generative artificial intelligence (1402), and can operate independently or integrated with other teams; • a model application process configured for pattern verification in order to identify preventive behavior, Information retrieval to provide contextual information for teams (1405) • an enhanced optimization process configured to optimize models to achieve the best preventive performance by performing hyperparameter adjustments, guiding team learning dynamically, influencing performance and maximizing the expected result, conditioned to the set of admissible actions determined by the evaluation of the Risk, Ethics and / or Review teams (1406); • an evaluation process designed to measure performance metrics, including accuracy, precision, recall and confusion matrix, serving as input to guide team decisions (1408).

2. Process, according to claim 1, characterized by creating the Ethics Team formed by artificial intelligence (1101), comprising a knowledge acquisition process from various data sources (1100), said team being configured to act in conjunction with decision-making processes, generate the MODIFIED SHEET (ARTICLE 19) humanized layer in the neural network, approval or disapproval of preventive methods (604), correction of preventive methods (606), through artificial intelligence techniques, including reinforcement optimization mechanism that applies restrictions through evaluation of regulatory compliance metrics.

3. Process, according to claim 1, characterized by creating a Risk Team formed by Artificial Intelligence (1200, 1202), comprising a knowledge acquisition process from various data sources (1201), configured to work in conjunction with decision-making processes, risk management, monitoring of atypical crimes and approval or rejection of risk plans (722), carrying out monitoring, controls and report management, through Artificial Intelligence techniques, selecting policies that maximize rewards associated with risk minimization, performing iterative self-adjustment for progressively increasing the accuracy of risk analysis and preventing the adoption of approaches that may generate unforeseen risks.

4. Process, according to claim 1, characterized by creating a Legal Advisor formed by Artificial Intelligence (1300, 1301, 1303), comprising a knowledge acquisition process from various data sources (1302), configured to act in conjunction with decision-making processes, generate the humanized layer in the neural network, prepare a technical legal opinion on the feasibility of preventive methods (708), propose decisions by reinforcement optimization, prevent actions that violate constitutional norms by penalizing during training and selection of actions, apply restrictions based on normative compliance metrics and iteratively adjust the model parameters to meet predefined constitutional thresholds.

5. Process, according to claims 1 and 2, characterized by eliminating discriminatory biases from existing data (200 to 210) through the action of the Ethics Team formed by Artificial Intelligence, comprising a process to detect and eliminate discriminatory biases using artificial intelligence techniques, configured to apply automatic penalties to the model outputs whenever biases are identified and iteratively adjust the decision parameters, in order to avoid discriminatory patterns in future predictions before they cause practical impact. MODIFIED SHEET (ARTICLE 19)6. Process, according to claims 1, 2 and 11, characterized by eliminating discriminatory biases from preventive methods created by Generative Artificial Intelligence (250 to 260) through the action of the Ethics Team formed by Artificial Intelligence, comprising a process to detect and eliminate discriminatory biases using artificial intelligence techniques, configured to apply automatic penalties to the model outputs whenever biases are identified and iteratively adjust the decision parameters, in order to avoid discriminatory patterns in future predictions before they cause practical impact.

7. A process, according to claims 1, 2, and 5, characterized by identifying preventive methods associated with crime prevention (300 to 308), prioritizing methods with the highest probability of effectiveness, iteratively adjusting recommendations based on automated impact simulations, and automatically preventing the selection of methods that may violate human rights and applicable legal norms.

8. A process, according to claim 1, characterized by generating knowledge from diverse data sources (500 to 503) to provide input to teams formed by Artificial Intelligence, configured to iteratively evaluate the relevance and reliability of the generated information, apply reinforcement optimization to identify potential legal conflicts, and, through evaluation of regulatory compliance metrics, iteratively adjust the model parameters to signal to the teams possible violations of legal norms.

9. A process, according to claim 1, characterized by generating knowledge from diverse data sources (400 to 403), to provide inputs to teams formed by Artificial Intelligence, configured to iteratively evaluate the relevance and reliability of the generated information, apply reinforcement optimization to identify potential legal conflicts, and, through evaluation of regulatory compliance metrics, iteratively adjust the model parameters in order to signal to the teams possible violations of Human Rights.

10. Process, according to claims 1, 2 and 7, characterized by identifying which preventive methods can be enhanced to prevent crimes (600 to 607), with the action of a team formed by Artificial Intelligence, applying reinforcement optimization that incorporates normative and ethical constraints directly into the decision function, MODIFIED SHEET (ARTICLE 19) and iteratively adjusting action strategies to maximize the effectiveness of preventive methods while ensuring compliance with legal and ethical principles.

11. Process, according to claims 1, 2, 7 and 10, characterized by creating new preventive methods for crime prevention (620 to 625) through Generative Artificial Intelligence with the action of a team formed by Artificial Intelligence, applying reinforcement optimization that incorporates normative and ethical constraints directly into the decision function, and iteratively adjusting the action strategies to maximize the effectiveness of the preventive methods while ensuring compliance with legal and ethical principles.

12. Process, according to claims 1, 2, 6, 8, 9, 10 and 11, characterized by verifying for violations of legislation and maintaining regulatory compliance (700 to 711), configured to support decision-making by Artificial Intelligence teams, which are adapted to apply enhanced optimization incorporating normative and ethical constraints directly into the decision function, iteratively adjusting hyperparameters and action strategies in order to ensure that all decisions comply with current legislation and principles of rights protection, operating as "checks and balances" mechanisms to proactively detect and prevent any legal impacts, providing automated, auditable and legally sound execution of complex decisions.

13. Process, according to claims 1, 2, 3 and 12, characterized by performing risk management (720 to 730) with the involvement of teams formed by Artificial Intelligence to support decision-making, configured to apply enhanced optimization directly to the decision function, with iterative adjustment of hyperparameters and action strategies for risk identification, monitoring and control, while implementing multi-agent interactions operating as prevention mechanisms and "checks and balances" designed to prevent the potentialization or escalation of risks throughout the decision-making process, providing automated, auditable and controlled execution of complex decisions.

14. Process, according to claims 1, 2, 3, 12 and 13, characterized by identifying new risks through Generative Artificial Intelligence performing risk management (740 to 751), with the action of teams formed by Artificial Intelligence to MODIFIED SHEET (ARTICLE 19) support for decision-making, configured to apply enhanced optimization directly to the decision function, with iterative adjustment of hyperparameters and action strategies for risk identification, monitoring, and control, while implementing multi-agent interactions operating as prevention mechanisms and "checks and balances" designed to prevent the potentialization or escalation of risks throughout the decision-making process, providing automated, auditable, and controlled execution of complex decisions.

15. Process, according to claims 1, 2, 8 and 9, characterized by performing knowledge updates from multiple data sources (800 to 803) provided to the Ethics Artificial Intelligence Team, in which the Team acts as an internal validation mechanism integrated into the decision-making process, evaluating parameters, data and results generated in order to automatically prevent any update or adjustment that violates predefined ethical criteria, so that the process provides knowledge application with preventive ethical control.

16. A process, according to claims 1, 2, 3 and 15, characterized by identifying, after updating legislation, whether a given conduct remains typified (820 to 825), being configured to work in conjunction with teams formed by Artificial Intelligence to support decision-making, applying enhanced optimization with direct incorporation of normative constraints in the decision function, and implementing multi-agent interactions that operate as "checks and balances" mechanisms intended to reassess the typification, providing automated, auditable and controlled execution of the decision-making process after normative changes.

17. A process, according to claims 1, 2, 3 and 15, characterized by identifying, after updating legislation, whether previously adopted preventive methods remain valid (840 to 845), configured to work in conjunction with teams formed by Artificial Intelligence to support decision-making, applying enhanced optimization with iterative adjustment of hyperparameters and action strategies to ensure that only preventive methods compatible with legislation and principles of rights protection remain active, while implementing multi-agent interactions operating as prevention mechanisms and "checks and balances" designed to block the maintenance or reactivation of incompatible methods. MODIFIED SHEET (ARTICLE 19) providing automated, auditable and controlled execution of the updating of preventive measures after regulatory changes.

18. A process, according to claims 1, 2, 3, 13, 14 and 17, characterized by performing continuous risk monitoring and control (860 to 868), configured to operate with teams formed by Artificial Intelligence to support decision-making, iteratively adjusting hyperparameters and action strategies for risk identification, containment, minimization and elimination, while implementing multi-agent interactions that operate as prevention mechanisms and "checks and balances", in which multi-agent control structurally interferes in the decision-making process through automatic parameter modification, restriction of decision alternatives or reconfiguration of action strategies, intended to prevent the materialization of risk actions before the generation of practical impact, providing automated, auditable and controlled execution of the decision-making process.

19. Process, according to claims 1, 2, 3, 5, 7, 8, 9, 10, 12 and 15, characterized by performing auditability monitoring and control (880 to 888), configured to record, track, correlate and validate steps of the decision-making process, integrating teams formed by Artificial Intelligence to support decision-making and generate explanatory records and structured information, in which auditability is applied in conjunction with enhanced optimization for iterative adjustment of parameters and action strategies based on verifiable criteria, allowing automatic reconstruction of the decision chain and independent validation of the system's behavior, providing transparency, reproducibility and structured technical control of the automated decision-making process.

20. Process, according to claims 1, 2 and 3, characterized by performing algorithmic management (900 to 906) with the participation of teams formed by Artificial Intelligence to support decision-making, applying enhanced optimization for iterative adjustment of parameters and action strategies based on performance and compliance criteria, while implementing multi-agent interactions that operate as prevention mechanisms and "checks and balances", in which multi-agent control interferes in the decision-making process through cross-validation and restriction of action alternatives, intended to prevent decisions that may compromise performance or MODIFIED SHEET (ARTICLE 19) compliance before generating practical impact, providing automated, auditable and controlled execution of the decision-making process.

21. Process, according to claims 1, 2, 3, 5 and 7, characterized by performing continuous monitoring and control of crime reduction after application of preventive methods (920 to 924), integrating teams formed by Artificial Intelligence for analysis, being configured to apply enhanced optimization with explicit incorporation of normative and ethical constraints directly in the decision function, with iterative adjustment of operational parameters to improve the effectiveness of preventive methods in concomitant normative compliance.

22. Process, according to claims 1 and 2, characterized by creating reports and managing reports (1000 to 1005), based on requests from an Ethics Team formed by Artificial Intelligence, maintaining the transparency and explainability of the processes, being configured to perform iterative adjustment of hyperparameters and action strategies to ensure that the reports reflect legal compliance, ethical principles and protection of rights. MODIFIED SHEET (ARTICLE 19)

Images