System and method for real-time intellectual property extraction and compliance in ai conversational environments

Abstract

A system (100) for intellectual property extraction and compliance in AI conversational environments includes a monitoring module (112) executed on a processor (104, 136) to analyze a dialogue stream. A state buffer (116) maintains a rolling context window of the dialogue stream. An NLP preprocessing pipeline (118) performs tokenization and chunking on the dialogue stream to generate discrete semantic units. A machine-learning classification engine (120) computes an extraction signal for the semantic units based at least in part on identifying linguistic patterns indicative of an inventive concept. A speaker role weighting algorithm applies a priority factor to human-authored input over AI-generated output in computing the extraction signal to isolate human-conceived subject matter. A syntactic claim generation engine (124) transforms the semantic units into a structured patent claim when the extraction signal satisfies a threshold. A compliance module (126) executes an inventorship validation sequence to verify human conception before generating a record in a secure disclosure record (130) associating the confirmed human contribution with the structured patent claim.

Description

SYSTEM AND METHOD FOR REAL-TIME INTELLECTUAL PROPERTY EXTRACTION AND COMPLIANCE IN AI CONVERSATIONAL ENVIRONMENTS

[0001] The present disclosure relates generally to automated legal documentation and natural language processing, and more specifically to a system for real-time intellectual property extraction and inventorship compliance within generative artificial intelligence conversational environments.

[0002] Traditional methods for identifying and capturing intellectual property rely heavily on retrospective manual disclosure processes. In these conventional workflows, an inventor must recognize a novel concept during or after a research and development phase and subsequently record that concept in a structured invention disclosure form. This process is inherently asynchronous and relies on the inventor’s subjective memory and initiative. While some automated tools exist for patent drafting, they are typically post-hoc utilities that require a user to provide a pre-defined technical summary or a completed software code base before any analysis can occur. These tools are incapable of identifying inventive steps as they emerge during the iterative process of ideation.

[0003] With the proliferation of generative artificial intelligence and large language model conversational interfaces, the pace of technical ideation has accelerated, creating a “velocity gap” between concept emergence and legal capture. Current conversational AI architectures are designed as ephemeral, session-based environments where data is processed for immediate response generation rather than long-term intellectual property management. Technical limitations in these platforms mean that inventive concepts generated during multi-turn dialogues are often lost once a session terminates or the context window shifts. Furthermore, existing natural language processing tools lack the specialized classification engines required to perform real-time semantic divergence analysis or to weight speaker roles in a way that isolates human-conceived inventive steps from machine-generated noise.

[0004] A significant technical deficiency in the state of the art is the inability to perform real-time granular attribution of inventive steps between human prompts and machine-generated outputs. Conventional systems cannot dynamically calculate the “claimability” of a dialogue stream or verify “human conception” and “non-routine contribution” according to rigorous legal standards during the act of creation. This results in a high degree of technical latency, where the transition from unstructured dialogue to structured, legally-defensible patent claims requires extensive manual labor and expert intervention. Existing monitoring tools are often fragile to schema changes in web interfaces and lack the low-latency interception capabilities necessary to maintain a rolling context window of complex, multi-turn technical exchanges. Therefore, there is a need for a system that can autonomously intercept, tokenize, and synthesize formal patent claims from human-AI dialogues while simultaneously executing a rule-based inventorship validation sequence to ensure legal compliance without disrupting the user’s creative workflow.DISCLOSURE OF THE INVENTION

[0005] To address the limitations of the prior art, such as the velocity gap between technical ideation and legal capture in generative AI environments, the present disclosure provides a system and method for real-time intellectual property extraction and compliance in AI conversational environments. The system fundamentally differs from conventional workflows by replacing retrospective, manual disclosure processes with an automated, real-time interception and analysis framework that operates concurrently with the creative process.

[0006] In one aspect, the system comprises a latency-optimized monitoring module (112), executed on a multi-core processor (136), configured to intercept a multi-turn dialogue stream between a human user and an artificial intelligence (AI) assistant via a transparent proxy layer (114). A state buffer (116) comprising high-speed memory is utilized to maintain a rolling context window of the multi-turn dialogue stream, ensuring that technical nuances are preserved across extended sessions. The system further includes an NLP preprocessing pipeline (118) configured to perform tokenization, lemmatization, part-of-speech (POS) tagging, and chunking on the intercepted dialogue stream to generate discrete semantic units. A machine-learning classification engine (120) is configured to dynamically calculate a claimability score for these units by evaluating functional language patterns and determining a concept novelty value based on a semantic divergence between a current dialogue turn and the rolling context window. To ensure legal defensibility, a speaker role weighting algorithm applies a mathematical priority factor to human-authored tokens over AI-generated tokens to isolate human-conceived inventive subject matter. When the claimability score exceeds a configurable threshold, a syntactic claim generation engine (124) transforms the discrete semantic units into a formal patent claim structure. A compliance module (126) then executes a rule-based inventorship validation sequence that computes a quantitative confidence score verifying human conception and non-routine contribution before committing the formal patent claim structure to a secure disclosure record (130).

[0007] The invention introduces several key innovations that overcome the deficiencies of session-based AI architectures. Through autonomous discovery, the system identifies inventive steps by observation of live interactions rather than requiring pre-defined technical summaries or manual triggers. It employs semantic abstraction to understand user intent and technical novelty from raw, unstructured dialogue, effectively filtering machine-generated noise from human innovation. By synthesizing formal legal structures from these interactions, the system performs cross-application synthesis, bridging the gap between ephemeral conversational tools and formal intellectual property management systems. Furthermore, the system provides personalization without extensive configuration by adapting its weighting algorithms to individual user patterns and specific speaker roles. This allows for the de novo creation of structured disclosure workflows that emerge naturally from the user’s existing creative habits rather than relying on rigid templates.

[0008] The technical advantages of this approach are significant. The system eliminates the manual workflow design burden typically associated with invention disclosures and discovers hidden efficiency opportunities by capturing concepts that might otherwise be lost to ephemeral context windows or session terminations. By handling heterogeneous application environments through a transparent proxy layer (114), the system reduces the technical expertise and effort required to maintain IP compliance during rapid R&D cycles. The continuous learning and refinement of the claimability thresholds ensure that the system adapts to the evolving technical complexity of the dialogue, providing a low-latency transition from unstructured ideation to legally-defensible documentation. These and other features, aspects, and advantages of the present disclosure will be better understood with reference to the following detailed description and accompanying drawings.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0010] illustrates a system architecture for real-time intellectual property extraction, in accordance with an embodiment of the present invention.

[0011] illustrates a real-time intellectual property extraction and compliance flow, in accordance with an embodiment of the present invention.

[0012] illustrates an intellectual property data object model, in accordance with an embodiment of the present invention.

[0013] illustrates a networked computing system component map, in accordance with an embodiment of the present invention.

[0014] illustrates a real-time extraction sequence, in accordance with an embodiment of the present invention.MODE FOR CARRYING OUT THE INVENTION

[0015] The Intellectual Property (IP) Extraction and Compliance System (100) is implemented across a distributed computing architecture comprising at least one User Device (102) communicatively coupled to a Networked Computing System (134) via a Network (108). The User Device (102) is a physical computing apparatus such as a desktop computer, laptop, or mobile device, including a Central Processing Unit (CPU) (104), a Memory (106), and a Non-Transitory Storage Medium (107) for storing an Operating System and application data. The User Device (102) executes an AI Conversational Interface (110), which may be a web-based interface or a native application, facilitating a multi-turn dialogue stream between a human user and an artificial intelligence assistant. A Latency-Optimized Monitoring Module (112) is instantiated within the Memory (106) of the User Device (102) or an intermediate edge server, comprising computer-executable instructions that, when executed by the Processor (104), establish a Transparent Proxy Layer (114). This Proxy Layer (114) is configured to intercept data packets or application-layer events and may be implemented as a Document Object Model (DOM) tracking script injected into the AI Conversational Interface (110), a native mobile hook for clipboard or keyboard capture, or an API middleware shim that intercepts traffic utilizing TCP / IP or TLS protocols.

[0016] To manage high-velocity data throughput without data loss, the System (100) includes a State Buffer (116) comprising a dedicated segment of high-speed memory, such as SRAM or DRAM, configured to maintain a rolling context window of the intercepted dialogue stream. This State Buffer (116) is communicatively coupled to an NLP Preprocessing Pipeline (118) executed by a Multi-Core Processor (136) within the Networked Computing System (134). The NLP Preprocessing Pipeline (118) comprises a series of computational stages configured to transform raw text into discrete semantic units through tokenization, lemmatization, part-of-speech (POS) tagging, and chunking. These discrete semantic units are then ingested by a Machine-Learning (ML) Classification Engine (120), which utilizes a Semantic Embedding Store (122) to calculate a claimability score. The ML Classification Engine (120) is configured to prioritize human-authored tokens over AI-generated tokens by applying a speaker role weighting algorithm stored in the Non-Transitory Storage Medium (138) of the Networked Computing System (134), ensuring that the resulting score reflects human-conceived inventive subject matter.

[0017] The System (100) further comprises a Syntactic Claim Generation Engine (124) and a Compliance Module (126), both implemented as software logic executed by the Multi-Core Processor (136). The Syntactic Claim Generation Engine (124) is configured to map the discrete semantic units to a formal patent claim structure, such as independent and dependent claims, stored in a Template Database (125). Simultaneously, the Compliance Module (126) executes a rule-based inventorship validation sequence, which compares the dialogue interactions against legal standards and jurisdictional requirements stored in a Jurisdictional Logic Controller (127). A Multi-Inventor Handler (128) is integrated within the Compliance Module (126) to manage collaborative sessions by tracking unique user identifiers, managing permissions, and facilitating digital signature collection.

[0018] Validated claims and associated inventor metadata are committed to a Secure Disclosure Record (130), which resides on a secure database server within the Networked Computing System (134). To ensure the legal defensibility and immutability of the captured IP, a Cryptographic Notary Interface (132) generates a unique cryptographic hash of the Secure Disclosure Record (130) and interfaces with an external timestamping service or distributed ledger via the Network (108). All internal components of the Networked Computing System (134) communicate via a high-speed System Bus (140) or internal local area network (LAN) utilizing secure protocols to ensure the real-time, low-latency transition from unstructured conversational ideation to structured, validated intellectual property documentation.

[0019] The System (100) manages data through a series of structured objects that evolve from raw telemetry to immutable legal records. The primary data structure ingested by the Transparent Proxy Layer (114) is the Raw Interaction Event Object, which is serialized to minimize latency. Each Raw Interaction Event Object comprises a unique Event_UUID, a high-precision Timestamp formatted with microsecond resolution, a Speaker_Role enumeration (e.g., HUMAN_USER, AI_ASSISTANT, SYSTEM_PROMPT), and a Raw_Payload containing the text string. This object further includes an Interface_Metadata field, which stores the Document Object Model (DOM) path or API endpoint identifier from the AI Conversational Interface (110) to provide structural context to the dialogue. These objects are temporarily persisted in the State Buffer (116) within a circular queue structure, allowing the System (100) to maintain a rolling context window of the most recent N-turns of dialogue for real-time analysis.

[0020] Upon transmission to the NLP Preprocessing Pipeline (118) via the System Bus (140), the raw text is transformed into a Semantic Unit Object. This intermediate data state is stored in the memory of the Networked Computing System (134) and structured to include a token array (e.g., a list of lemmatized strings) and a part-of-speech mapping. Crucially, this object incorporates a vector embedding field, which stores a high-dimensional array generated by the Semantic Embedding Store (122). This vector representation allows the ML Classification Engine (120) to perform mathematical distance calculations against known prior art or functional patterns. The ML Classification Engine (120) then appends a metadata header to the Semantic Unit Object, generating an extraction signal (also referred to herein as a structured claimability signal). While this extraction signal may be embodied as a quantitative claimability score (e.g., a float between 0.0 and 1.0), it may also comprise any structured output, boolean flag, or semantic tag that indicates the presence of patentable subject matter derived from the functional language and semantic divergence calculations.

[0021] When the extraction signal satisfies a threshold defined in the Non-Transitory Storage Medium (138), the Syntactic Claim Generation Engine (124) is triggered to map the semantic units to a formal legal structure. The Syntactic Claim Generation Engine (124) queries the Template Database (125) to instantiate a hierarchical claim structure based on syntactic dependency trees. Each node in this hierarchical structure represents a claim element, comprising fields for the element type (e.g., preamble, transitional phrase, limitation), the synthesized technical prose, and a dependency link (referencing a parent node for dependent claims). This structure ensures that the unstructured conversational data is mapped to a rigid, legally compliant format suitable for patent filing.

[0022] Simultaneously, the Compliance Module (126) generates a Validation State Object, which is stored in the Jurisdictional Logic Controller (127). This object tracks the legal defensibility of the generated IP and includes fields for Jurisdiction_ID (e.g., USPTO, EPO), an Inventorship_Confidence_Score, and a Multi_Inventor_Manifest. The Multi_Inventor_Manifest is an array of sub-objects managed by the Multi-Inventor Handler (128), where each sub-object contains a User_ID, a Contribution_Timestamp, and a Digital_Signature_Blob. This manifest ensures that every segment of the hierarchical claim structure is programmatically linked to a specific human contributor, satisfying the requirements for human conception.

[0023] The final data state is the Secure Disclosure Record (130), which is committed to a relational database or an encrypted document store within the Networked Computing System (134). The Secure Disclosure Record (130) serves as the terminal executable output and comprises the complete hierarchical claim structure, the associated Validation State Object, and a Notary_Header. The Notary_Header, generated by the Cryptographic Notary Interface (132), contains a cryptographic hash of the entire record, a Merkle Root if part of a distributed ledger transaction, and a Trusted_Timestamp. For transmission across the Network (108), these structures are encapsulated in TLS-encrypted packets, ensuring that sensitive intellectual property remains secure in transit between the User Device (102) and the Networked Computing System (134). Schema evolution is managed through version-stamping within each object header, allowing the System (100) to maintain backward compatibility with legacy dialogue formats while updating jurisdictional logic in the Controller (127).

[0024] During normal system operation, a human user engages in a multi-turn dialogue with an artificial intelligence assistant via the AI Conversational Interface (110) executed on the User Device (102). As the dialogue progresses, the Latency-Optimized Monitoring Module (112) instantiates the Transparent Proxy Layer (114) to intercept application-layer events and data packets in real-time. The Transparent Proxy Layer (114) captures text strings and associated metadata, such as Document Object Model (DOM) paths or API endpoints, and serializes this data into a Raw Interaction Event Object. Each object is assigned a unique Event_UUID, a Speaker_Role enumeration, and a high-precision Timestamp. The Raw Interaction Event Object is immediately persisted within the State Buffer (116), which utilizes a circular queue in high-speed memory (106) to maintain a rolling context window of the most recent dialogue turns.

[0025] Responsive to the arrival of new telemetry in the buffer, the System (100) transmits the Raw Interaction Event Object across the Network (108) to the Networked Computing System (134). Upon receipt, the NLP Preprocessing Pipeline (118) ingests the raw payload and executes a series of computational stages, including tokenization, lemmatization, and part-of-speech (POS) tagging, to transform the unstructured text into a structured Semantic Unit Object. This object includes a Token_Array and a Dependency_Graph_Pointer, which are stored in the memory of the Networked Computing System (134) for immediate processing. The ML Classification Engine (120) then retrieves the Semantic Unit Object and interfaces with the Semantic Embedding Store (122) to generate a high-dimensional Vector_Embedding_Field. Using this vector representation, the ML Classification Engine (120) performs mathematical distance calculations against known functional patterns to determine a Novelty_Index and a functional Claimability_Score. Simultaneously, the ML Classification Engine (120) applies a speaker role weighting algorithm to the Token_Array, referencing the Speaker_Role enumeration to calculate a Human_Contribution_Ratio that prioritizes human-authored inventive concepts over AI-generated suggestions.

[0026] When the Claimability_Score exceeds a predefined threshold stored in the Non-Transitory Storage Medium (138), the Syntactic Claim Generation Engine (124) is triggered to map the semantic units to a formal legal structure. The Syntactic Claim Generation Engine (124) queries the Template Database (125) to instantiate a hierarchical claim structure, where each node represents a Claim_Element_Object. The Syntactic Claim Generation Engine (124) populates these nodes with synthesized technical prose, establishing logical Dependency_Links between independent and dependent claim elements to ensure antecedent basis and structural integrity. Concurrently, the Compliance Module (126) initiates a rule-based inventorship validation sequence by querying the Jurisdictional Logic Controller (127). The Compliance Module (126) generates a Validation State Object that evaluates the dialogue interactions against specific jurisdictional requirements to produce an Inventorship_Confidence_Score. If multiple users are detected in the dialogue stream, the Multi-Inventor Handler (128) updates a Multi_Inventor_Manifest within the Validation State Object, tracking unique User_IDs and Contribution_Timestamps while facilitating the collection of Digital_Signature_Blobs for each identified contributor.

[0027] Subsequent to the validation of the claim structure and inventorship, the System (100) commits the complete hierarchical claim structure and the associated Validation State Object to the Secure Disclosure Record (130). To ensure the legal defensibility and immutability of the captured intellectual property, the Cryptographic Notary Interface (132) computes a cryptographic hash of the entire record and appends a Notary_Header containing a Trusted_Timestamp. This finalized Secure Disclosure Record (130) is then persisted in an encrypted database within the Networked Computing System (134) or interfaced with an external distributed ledger via the Network (108). Throughout this sequence, the System (100) maintains backward compatibility through version-stamping within each object header, allowing the Jurisdictional Logic Controller (127) to update its internal rules without disrupting the real-time extraction of intellectual property from the continuous dialogue stream.

[0028] One of ordinary skill in the art will appreciate that the specific hardware, software, and network configurations described in the foregoing embodiments are provided for illustrative purposes and are not intended to limit the scope of the disclosure. The system and method may be implemented using any number of technical variations that achieve the same inventive function of real-time intellectual property extraction and compliance validation. While the networked computing system is described as a cloud-based server, the invention may be implemented on edge devices, a local intranet, a private enterprise data center, a peer-to-peer network, or a blockchain-based distributed ledger. The processing need not occur on a single device; for example, the monitoring and buffering steps may occur on a mobile device or local workstation, while the machine learning classification and claim generation occur on a remote server or a distributed computing cluster.

[0029] Any processor configured to perform the described operations is within the scope of this disclosure, regardless of architecture, including but not limited to x86, ARM, RISC-V, Graphics Processing Units (GPUs), Tensor Processing Units (TPUs), Field Programmable Gate Arrays (FPGAs), or Application-Specific Integrated Circuits (ASICs). Furthermore, the software logic and modules described herein may be implemented in any suitable programming language or environment, such as Python, Java, C++, Go, Rust, or JavaScript, and may be deployed as microservices, containerized applications, or monolithic software packages.

[0030] The monitoring of the AI conversational interface may be achieved through various technical means beyond the described transparent proxy layer. Alternative embodiments may utilize operating system-level hooks, keyboard and clipboard capture drivers, network-level packet inspection, browser extensions, or direct integration via native Software Development Kits (SDKs) and Application Programming Interfaces (APIs) such as REST, GraphQL, or gRPC. The capture of data is not limited to UTF-8 text and may include voice-to-text streams, image-based screen scraping via Optical Character Recognition (OCR), or structured data payloads from integrated development environments (IDEs).

[0031] While a machine learning classification engine utilizing neural networks is described, the logic for calculating claimability scores and novelty indices may be implemented via heuristic rules, decision trees, support vector machines, Bayesian networks, or regression analysis. The NLP preprocessing pipeline may utilize various transformer-based architectures, such as BERT or GPT variants, or may rely on traditional statistical sequence mining and linguistic pattern matching. Reference to a relational database for the secure disclosure record or template database is exemplary; alternative embodiments may utilize key-value stores, document-oriented NoSQL databases, graph databases for tracking complex inventor relationships, or vector indexes for high-dimensional semantic search.

[0032] The cryptographic notary interface may employ any standard or proprietary hashing algorithm, such as SHA-256, SHA-3, or BLAKE2, and may interface with various immutability layers, including centralized trusted timestamping authorities or decentralized protocols like Ethereum, Hyperledger, or other distributed ledgers. One of ordinary skill in the art will further recognize that the order of operations may be modified, and certain steps may be performed concurrently or omitted without departing from the core inventive concept of transforming unstructured conversational ideation into structured, validated intellectual property documentation. The jurisdictional logic controller may be updated dynamically via remote policy pushes, and the multi-inventor handler may support various authentication methods, including biometric verification, multi-factor authentication (MFA), or decentralized identifiers (DIDs). These and other technical equivalents are intended to be encompassed within the scope of the present disclosure.

[0033] The intellectual property extraction and compliance system described herein is not limited to the specific embodiments of patent claim generation from AI-human dialogues but extends to a wide array of industrial and functional applications where unstructured data must be transformed into structured, compliant documentation. In various alternative embodiments, the core innovation of real-time event mapping and semantic scoring is applied across diverse verticals to solve unique data integrity and compliance challenges.

[0034] In healthcare implementations, the system is configured to intercept dialogues between medical professionals and clinical AI assistants to generate structured patient records, clinical trial protocols, or diagnostic care plans. In this context, the NLP preprocessing pipeline (118) is optimized for domain-specific medical ontologies and adherence to healthcare data standards. The machine-learning classification engine (120) evaluates the “clinical significance” of a dialogue turn rather than “claimability,” ensuring that human-conceived treatment decisions are prioritized and cryptographically logged for medical privacy compliance and medical malpractice defense.

[0035] In the finance and fintech sectors, the system is utilized to monitor interactions between financial analysts and AI-driven market intelligence tools. The system transforms unstructured investment brainstorming into structured ledger entries, compliance reports, or algorithmic trading strategies. The jurisdictional logic controller (127) is updated with appropriate domain-specific regulatory requirements, and the state buffer (116) maintains a context window of market data to validate the human analyst’s rationale behind a specific trade or investment recommendation, providing an immutable audit trail for regulatory oversight.

[0036] Within the logistics and supply chain industry, the invention is deployed to capture real-time coordination between logistics managers and AI routing assistants. The system extracts supply chain optimizations, such as novel route configurations or warehouse management protocols, and transforms them into structured shipping manifests or operational procedures. The multi-inventor handler (128) manages collaborative sessions between global teams, tracking unique identifiers across different time zones and jurisdictions to ensure that logistical innovations are properly attributed and documented within a secure disclosure record (130).

[0037] In cybersecurity applications, the system functions as a real-time threat intelligence capture layer. As security analysts interact with AI-powered security platforms, the system intercepts the dialogue to generate structured threat detection logs or incident response playbooks. The ML classification engine (120) identifies novel attack patterns or defensive strategies, while the cryptographic notary interface (132) ensures that the resulting security documentation is tamper-proof, providing a legally defensible record of the organization’s incident response efficacy.

[0038] Functional expansion of the system includes horizontal applications across various corporate departments. In Human Resources, the system is configured to automate employee onboarding and performance review workflows by extracting structured goals and policy agreements from conversational interactions. In software development environments, the system intercepts developer-AI dialogues to generate deployment orchestrations, infrastructure templates, or software architecture documentation, ensuring that technical debt is minimized through real-time documentation. In legal departments, the system facilitates contract review automation by mapping conversational negotiations to formal legal clauses and verifying them against a jurisdictional database of case law and statutory requirements.

[0039] The invention further scales across different user tiers and architectural environments. In a consumer-facing scale, the system operates in a single-user mode on a personal device, such as a smartphone or tablet, providing individual creators with a lightweight tool for capturing and notarizing personal ideation. In an enterprise-scale variation, the system is implemented as a multi-tenant SaaS architecture serving thousands of concurrent organizations, utilizing distributed computing clusters and high-availability database shards to manage massive volumes of multi-turn dialogue streams. The system may be deployed on-premise for high-security environments, at the network edge for low-latency industrial applications, or across a decentralized peer-to-peer network to ensure maximum data redundancy.

[0040] The technical implementation of the system is agnostic to the specific programming languages or hardware architectures utilized. While Python and Java may be used for the NLP and ML components, low-level languages such as C++ or Rust may be employed for the latency-optimized monitoring module (112) and state buffer (116) to maximize throughput. Data storage may vary from relational SQL databases for structured metadata to NoSQL document stores for raw payloads and vector databases for high-dimensional semantic embeddings. The cryptographic notary interface (132) may utilize various hashing algorithms beyond SHA-256 and may interface with public or private blockchains, centralized timestamping authorities, or hardware security modules (HSMs).

[0041] The specific configurations, choice of materials, and the size and shape of various elements can be varied according to particular design specifications or functional requirements. The detailed description is to be construed as exemplary only and not as limiting the invention.

[0042] The system and method for real-time intellectual property extraction and compliance disclosed herein are susceptible of industrial application and can be manufactured and used in a broad range of technology industries. The system is implementable on commodity computer hardware, mobile devices, edge servers, and distributed cloud computing infrastructure using standard software development tools and programming languages, enabling manufacturers of enterprise software, research and development platforms, and intellectual property management systems to integrate the invention into existing product lines and service offerings.

[0043] Industries that may utilize the disclosed system include, without limitation, enterprise research and development organizations seeking to capture inventive subject matter generated during human-AI collaboration; intellectual property law firms and patent management services seeking to automate invention disclosure workflows; healthcare providers seeking to document clinical decision making in compliance with medical privacy regulations; financial institutions seeking to generate auditable records of human-authored investment rationale; logistics and supply chain operators seeking to document novel operational processes; cybersecurity operations seeking tamper-evident records of incident response activity; and software development organizations seeking automated documentation of architectural and infrastructure decisions. The invention may be offered as a software-as-a-service subscription, an on-premise enterprise deployment, an embedded library within third-party conversational AI platforms, or a browser extension for end-user installation, thereby demonstrating broad industrial applicability across the information technology and professional services sectors.

Claims

A system for real-time intellectual property extraction and compliance, comprising at least one processor (104, 136) and a memory (106, 138) storing instructions executable by the at least one processor to analyze a dialogue stream between a human user and an artificial intelligence (AI) system and to generate a record of inventive subject matter derived therefrom,characterized in that the instructions, when executed by the at least one processor (104, 136), further configure the system to:(a) compute an extraction signal for a segment of the dialogue stream based at least in part on identifying linguistic patterns indicative of an inventive concept;(b) prioritize human-authored input over AI-generated output in computing the extraction signal;(c) generate, when the extraction signal satisfies a threshold, a structured patent claim based on the segment using a syntactic claim generation engine (124);(d) execute, by a compliance module (126), an inventorship validation sequence to prompt the human user for confirmation of human contribution to the structured patent claim; and(e) generate, in a secure disclosure record (130), a record associating the confirmed human contribution with the structured patent claim.The system of claim 1, wherein the instructions further configure the system to maintain a rolling context window of the dialogue stream in a state buffer (116), and wherein the extraction signal is further computed by evaluating a semantic divergence between the segment and the rolling context window.The system of claim 1, wherein prioritizing human-authored input comprises applying, by a machine-learning classification engine (120), a mathematical speaker role weighting to discrete human-authored tokens parsed from the dialogue stream.The system of claim 1, wherein the instructions further configure the system to intercept the dialogue stream utilizing at least one deployment architecture selected from a group consisting of: a browser extension configured to inject tracking scripts into a Document Object Model (DOM) via a transparent proxy layer (114), a native mobile application utilizing clipboard listening or speech-to-text hooks, and an API middleware proxy.The system of claim 1, wherein generating the structured patent claim comprises generating, by the syntactic claim generation engine (124) using a template database (125), a claim family draft encompassing a system claim, a method claim, an apparatus claim, and a computer-readable medium claim.The system of claim 1, wherein the inventorship validation sequence comprises a prompt engine configured to administer structured questions based on legal inventorship standards, stored in a jurisdictional logic controller (127), to compute a quantitative confidence score verifying non-routine human contribution to the structured patent claim.A method for real-time intellectual property extraction and compliance, comprising analyzing, by a computing system, a dialogue stream between a human user and an artificial intelligence (AI) system, and generating a record of inventive subject matter derived therefrom,characterized by:(a) computing an extraction signal for a segment of the dialogue stream based at least in part on identifying linguistic patterns indicative of an inventive concept;(b) prioritizing human-authored input over AI-generated output in computing the extraction signal;(c) generating, when the extraction signal satisfies a threshold, a structured patent claim based on the segment;(d) executing an inventorship validation sequence to prompt the human user for confirmation of human contribution to the structured patent claim; and(e) generating a record associating the confirmed human contribution with the structured patent claim.The method of claim 7, further comprising maintaining a rolling context window of the dialogue stream in a state buffer (116), wherein computing the extraction signal further comprises evaluating a semantic divergence between the segment and the rolling context window.The method of claim 7, wherein prioritizing human-authored input comprises applying a mathematical speaker role weighting to discrete human-authored tokens parsed from the dialogue stream by a machine-learning classification engine (120).The method of claim 7, wherein executing the inventorship validation sequence comprises administering, by a compliance module (126), structured questions based on legal inventorship standards to compute a quantitative confidence score verifying non-routine human contribution.A non-transitory computer-readable medium (107, 138) storing instructions that, when executed by at least one processor (104, 136), perform a method for real-time intellectual property extraction and compliance comprising analyzing a dialogue stream between a human user and an artificial intelligence (AI) system and generating a record of inventive subject matter derived therefrom,characterized in that the method further comprises:(a) computing an extraction signal for a segment of the dialogue stream based at least in part on identifying linguistic patterns indicative of an inventive concept;(b) prioritizing human-authored input over AI-generated output in computing the extraction signal;(c) generating, when the extraction signal satisfies a threshold, a structured patent claim based on the segment;(d) executing an inventorship validation sequence to prompt the human user for confirmation of human contribution to the structured patent claim; and(e) generating, in a secure disclosure record (130), a record associating the confirmed human contribution with the structured patent claim.

Images