Management method and system for generated content of intelligent agent

Abstract

The invention discloses a management method and system for generated content of an intelligent agent. The method comprises the following steps: in response to generation condition information of an input target agent, obtaining an information abstract corresponding to the generation condition information; acquiring an agent identifier of the target agent; in response to the target agent, determining one or more target models based on the generation condition information, and obtaining a model identifier of each target model in the one or more target models; and in response to the target agent, outputting a generation content based on the one or more target models, determining a causal identification chain of the generation content based on the information abstract, the agent identifier and the model identifier of each target model in the one or more target models, and storing the causal identification chain in a distributed account book or a trusted evidence storage system.

Description

Technical Field

[0001] This invention relates to the field of computers, and more particularly to a method and system for managing the generated content of intelligent agents. Background Technology

[0002] With the rapid development of large language models, multi-agent systems and generative artificial intelligence, artificial intelligence systems are gradually showing characteristics such as complex data sources, frequent model iterations, huge scale of generated content, and increased autonomy of agents. Summary of the Invention

[0003] According to an embodiment of the present invention, a method for managing generated content for an intelligent agent includes: in response to inputting generation condition information of a target intelligent agent, obtaining a message digest corresponding to the generation condition information; obtaining an intelligent agent identifier for the target intelligent agent; in response to the target intelligent agent determining one or more target models based on the generation condition information, obtaining a model identifier for each of the one or more target models; and in response to the target intelligent agent outputting generated content based on one or more target models, determining a causal identification chain of the generated content based on the message digest, the intelligent agent identifier, and the model identifier for each of the one or more target models, and storing the causal identification chain in a distributed ledger or a trusted evidence storage system.

[0004] A management system for generating content for an intelligent agent according to an embodiment of the present invention includes: an information digest determination module configured to obtain an information digest corresponding to the generation condition information in response to input generation condition information of a target intelligent agent; an intelligent agent identifier determination module configured to obtain an intelligent agent identifier of the target intelligent agent; a model identifier determination module configured to obtain a model identifier of each of the one or more target models in response to the target intelligent agent determining one or more target models based on the generation condition information; and a causal identification chain determination module configured to determine a causal identification chain of the generated content based on the information digest, the intelligent agent identifier, and the model identifier of each of the one or more target models in response to the target intelligent agent outputting generated content based on one or more target models, and store the causal identification chain in a distributed ledger or a trusted evidence storage system.

[0005] According to an embodiment of the present invention, a computer-readable storage medium stores computer-executable instructions thereon, wherein, when executed by a processor, these computer-executable instructions cause the processor to perform the above-described method for managing the generated content of an intelligent agent.

[0006] A computer program product according to an embodiment of the present invention includes computer-executable instructions, wherein, when executed by a processor, these computer-executable instructions cause the processor to perform the above-described method for managing the generated content of an intelligent agent. Attached Figure Description

[0007] The invention can be better understood from the following description of specific embodiments of the invention in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a method for managing generated content for intelligent agents according to an embodiment of the present invention is shown.

[0008] Figure 2 A schematic diagram illustrating relevant information about a method for managing generated content for intelligent agents according to an embodiment of the present invention is shown.

[0009] Figure 3 Another flowchart illustrating a method for managing generated content for intelligent agents according to an embodiment of the present invention is shown.

[0010] Figure 4 A schematic block diagram of a management system for generating content for intelligent agents according to an embodiment of the present invention is shown.

[0011] Figure 5 A schematic diagram of a computer system is shown that can implement a method and apparatus for managing generated content for intelligent agents according to embodiments of the present invention. Detailed Implementation

[0012] The features and exemplary embodiments of various aspects of the present invention will now be described in detail. Numerous specific details are set forth in the following detailed description to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention may be practiced without requiring some of these specific details. The following description of embodiments is merely intended to provide a better understanding of the invention by illustrating examples of the invention. The invention is by no means limited to any specific configurations and algorithms presented below, but covers any modifications, substitutions, and improvements to elements, components, and algorithms without departing from the spirit of the invention. Well-known structures and techniques are not shown in the drawings and the following description in order to avoid unnecessarily obscuring the invention.

[0013] AI-generated content faces challenges in tracing its origins and ensuring sufficient traceability due to the complexity of its data sources and frequent model iterations. Furthermore, the massive scale of generated content and the increased autonomy of intelligent agents present technical challenges to content regulation, including real-time performance, coverage, and adaptability, further increasing the technical complexity of systematic governance.

[0014] According to embodiments of the present invention, during the process of a target intelligent agent outputting generated content based on input generation condition information, a causal identification chain is generated based on the information digest of the generation condition information, the agent identifier of the target intelligent agent, and the model identifier of the target model used by the target intelligent agent. This causal identification chain is then stored in a distributed ledger or a trusted evidence storage system. This embodiment of the present invention achieves automatic recording of the causal identification chain. Once stored in a distributed ledger or trusted evidence storage system, the causal identification chain cannot be tampered with and can be used for content traceability and accountability, providing technical evidence support for content supervision.

[0015] Figure 1 A flowchart illustrating a method for managing generated content for intelligent agents according to an embodiment of the present invention is shown. Figure 1 As shown, the method 100 for managing generated content for an intelligent agent includes steps S101-S104. S101: In response to inputting generation condition information of the target intelligent agent, obtain a message digest corresponding to the generation condition information. S102: Obtain the agent identifier of the target intelligent agent. S103: In response to the target intelligent agent determining one or more target models based on the generation condition information, obtain the model identifier of each of the one or more target models. S104: In response to the target intelligent agent outputting generated content based on one or more target models, determine the causal identification chain of the generated content based on the message digest, the agent identifier, and the model identifier of each of the one or more target models, and store the causal identification chain in a distributed ledger or trusted evidence storage system.

[0016] In the method according to embodiments of the present invention, the working process of the AI ​​agent includes: receiving input generation condition information, selecting one or more target models based on the generation condition information, and outputting generated content corresponding to the generation condition information based on the one or more target models. Based on the working process of the AI ​​agent, the information and objects involved are identified, an associated causal identification chain is established, and stored in an immutable location. According to the causal identification chain of the generated content, the target AI agent, target model, and generation condition information corresponding to the generated content can be traced, which is beneficial for accountability and content supervision.

[0017] In some embodiments, the information digest of the generated condition information can be obtained by performing a hash operation on the generated condition information.

[0018] In the method according to embodiments of the present invention, what is stored in the distributed ledger or trusted evidence storage system can be a complete causal identification chain or a summary state of the causal identification chain. Therefore, in some embodiments, storing the causal identification chain in the distributed ledger or trusted evidence storage system may include: storing a state summary of the causal identification chain in the distributed ledger or trusted evidence storage system. The state summary of the causal identification chain can be obtained by performing a hash operation on the causal identification chain.

[0019] In some embodiments, the factors for determining the causal identification chain may further include the output time of the target agent's output content. Accordingly, determining the causal identification chain may include: determining the causal identification chain based on the output time of the target agent's output content, the information digest, the agent identifier, and the model identifier of each of the one or more target models.

[0020] In some embodiments, the agent identifier may include: the scope of action permissions for the target agent, and a responsibility attribution identifier for the owner of the target agent. The scope of action permissions includes all operations permitted to be performed by the agent, typically determined based on policy constraints and compliance requirements. The responsibility attribution identifier is a unique identifier for the owner of the agent, who is responsible for all actions of the target agent. The agent identifier may also include an identity token and a revocation rule. The identity token represents the identity of the target agent. The revocation rule specifies under what circumstances the target agent will be revoked.

[0021] In some embodiments, the model identifier of any target model may include: a dataset identifier for each of the one or more training datasets used to train the target model. In some embodiments, the dataset identifier of any training dataset may include: data source information, data ownership information, and data usage constraint information. The data source information indicates the source from which each data point in the dataset was collected. The data ownership information indicates the owner of the dataset. The data usage constraint information, also known as usage license information, includes the permitted uses of the dataset, the validity period of use, and the region of use.

[0022] In some embodiments, the current model identifier of any target model may further include: at least one of the resources, methods, and process features used to train the target model; the architecture of the target model; and descriptive information for describing the target model. In the model identifier, the resources used to train the target model refer to at least one of the software or hardware resources used during target model training. Similarly, the methods used to train the target model refer to the algorithms used during target model training, such as supervised learning, unsupervised learning, transfer learning, federated learning, etc. The process features used to train the target model include metadata and key metrics generated during the dynamic process of model training that reflect the training state, behavior, and environment. The descriptive information of the target model is used to describe the functional purpose or other aspects of the target model. The dataset, resources, methods, and architecture of the target model used for training may be adjusted according to the task objectives and resource constraints of the target model. The method in this embodiment does not set or modify the above information; it only organizes and records this information to determine the model identifier.

[0023] The model used by the intelligent agent may generate information different from that generated during the training process during application. Therefore, the model identifier in this embodiment of the invention includes not only information from the training process and the results after training, but also information exhibited during the application process. This can be achieved by periodically detecting the target model in the application. Therefore, in some embodiments, the model identifier of any target model may also include: performance metrics and policy compliance evaluation results added after periodically detecting the target model; and abnormal behavior metrics added after periodically detecting the target model and detecting anomalies in the target model. Specifically, the newly added performance metrics, policy compliance evaluation results, and abnormal behavior metrics are added to the original model identifier. The results of each detection, including performance metrics and policy compliance evaluation results, are appended to the original model identifier. Abnormal behavior metrics are only appended to the original model identifier when anomalies are detected in the target model.

[0024] In a specific example, the dataset identifier for a training dataset can be represented as a JSON structure like the one shown below: DP-ID:{ source:"licensed_news_dataset_v3", usage:"training_allowed", jurisdiction:"EU", *expiry:"2027-12-31"* } The DP-ID (Data Provenance ID) is the structure name of this dataset identifier. `source: "licensed_news_dataset_v3"` indicates the data source information. `usage: "training_allowed"` indicates that the dataset is permitted for model training. `jurisdiction: "EU"` indicates that the dataset is permitted to be used in the European Union. `expiry: "2027-12-31"` indicates that the dataset's validity period ends on December 31, 2027.

[0025] In a specific example, the model identifier of a target model can be represented as a JSON structure like the one shown below: MC-ID: { base_model: "LLM-7B", training_data_refs: [DP-1023, DP-1178], capabilities: ["text_generation", "summarization"], risk_level: "medium" } Here, MC-ID (Model Capability ID) is the structure name of this model identifier. `base_model: "LLM-7B"` indicates the architecture of the target model. `training_data_refs: [DP-1023, DP-1178]` represents the training dataset used to train this target model. `capabilities: ["text_generation", "summarization"]` represents descriptive information used to describe the target model, describing its purpose. The model identifier can also include the risk level of the target model, such as `risk_level: "medium"` indicating that the risk level of the target model is medium.

[0026] In a specific example, the agent identifier of a target agent can be represented as a JSON structure like the one shown below: AS-ID: { agent_type: "customer_support_bot", permissions: ["respond", "summarize"], max_autonomy: "bounded", owner: "Company_X" } Here, AS-ID (Agent Subject ID) is the structure name of this agent identifier. `agent_type: "customer_support_bot"` indicates that the agent type is user support, i.e., customer service and support. `permissions: ["respond", "summarize"]` indicates that the agent's action permissions include responding and summarizing. `max_autonomy: "bounded"` indicates that the highest level of autonomous action for this agent is restricted autonomy; the agent is designed to exercise autonomy within preset rules, defined scopes, or human oversight frameworks. `owner: "Company_X"` indicates that the responsibility identifier for the agent's owner is Company_X.

[0027] In one specific embodiment, a causal identifier chain for generating content can be represented as a JSON structure as shown below: CC-ID: { content_hash: "0xA912F...", model_ref: MC-7781, agent_ref: AS-4402, policy_state: "compliant" } Here, CC-ID (Content Creation ID) is the structure name of this causal identification chain. `content_hash: "0xA912F..."` represents the message digest corresponding to the generation condition information, which can be represented as a hash code. `model_ref: MC-7781` represents the model identifier of the target model involved. `agent_ref: AS-4402` represents the agent identifier of the target agent involved. `policy_state: "compliant"` indicates that the compliance check result is compliant with regulations.

[0028] In the method according to embodiments of the present invention, the agent identifier, model identifier, dataset identifier, and causal identification chain may include elements not shown in the above examples, or may represent forms different from those shown in the above examples. The generation of the agent identifier, model identifier, and dataset identifier typically occurs before the target agent receives the generation condition information, and the generation or update of these identifiers typically follows the creation or update of the corresponding target agent, target model, and training dataset.

[0029] Figure 2The diagram illustrates relevant information about a method for managing generated content for intelligent agents according to an embodiment of the present invention. Example 2 uses a target model and a training dataset as an example. Figure 2 As shown, the target agent receives the input generation condition information, selects a target model based on the generation condition information, and outputs the generated content corresponding to the generation condition information based on the target model. For example... Figure 2 As shown, the causal identification chain can be determined based on the information digest of the generation condition information, the agent identifier of the target intelligent agent, the model identifier of the target model, and the dataset identifier of the training dataset. In the method according to embodiments of the present invention, the causal identification chain cryptographically or logically binds the logical relationship between the generated content, the target intelligent agent, the target model, the training dataset, and the generation condition information. This can be used to trace and hold accountable the entire process of content generation, thereby solving the intractable identity collapse problem in AI governance. Specifically, the causal identification chain achieves content traceability, allows for responsibility allocation and accountability based on the agent identifier and model identifier, and allows for tracing the copyright and authorization status of the training dataset and analyzing data bias based on the dataset identifier associated with the model identifier.

[0030] Figure 3 Another flowchart illustrating a method for managing generated content for intelligent agents according to an embodiment of the present invention is shown. Figure 3 As shown, the method 100 in this embodiment may further include: S105: analyzing whether the target agent is compatible with one or more target models based on the agent identifier and the model identifier of each of the one or more target models; S106: if the target agent is incompatible with at least one of the one or more target models, instructing the target agent to stop outputting generated content. If the target agent is compatible with all target models, the target agent will automatically output generated content.

[0031] In some embodiments, analyzing the compatibility of a target agent with a target model includes analyzing whether the permitted scopes of the target agent, the target model, and the training dataset used to train the target model are consistent. For example, the dataset identifier of the training dataset indicates that the training dataset is EU financial transaction data, which is only permitted for compliance monitoring and not for trading; the model identifier of the target model indicates that the target model is a model used for "high-frequency trading strategies"; and the agent identifier of the target agent indicates that the target agent is authorized to rebalance portfolios in the Asian market. The analysis process includes: Based on the model identifier of the target model and the corresponding dataset identifier of the training dataset, it can be seen that the training dataset is restricted to compliance monitoring, while the target model is designed for trading. Since the intended use of the target model conflicts with the scope of use allowed by the training dataset, the two are incompatible. Based on the model identifier of the target model and the agent identifier of the target agent, it can be seen that the target model is used for high-frequency trading, but the target agent is only authorized to perform rebalancing operations in Asia. Since the scope of authority of the target agent does not include high-frequency trading, the two are incompatible.

[0032] In the method according to this embodiment, the causal identification chain can be verified to verify the existence of the causal identification chain and its elements. In some embodiments, the management method may further include: verifying the causal identification chain using hash proofs and ledger checks. Hash proofs are used to transform raw data into a fixed-length hash value, serving as a unique "fingerprint" and commitment of the data, while ledger checks are used to prove that a hash commitment has been recorded in an immutable distributed ledger (such as a blockchain).

[0033] Furthermore, the verification of causal identification chains can also employ zero-knowledge proofs to further prove certain properties of the data (such as range and format) without revealing the data itself. In some embodiments, verifying causal identification chains may include using hash proofs, ledger checks, and zero-knowledge proofs to verify the causal identification chains.

[0034] In some embodiments, the agent identifier of the target agent can be configured to be revocable. When the agent identifier of the target agent is configured to be unrevocable, the target agent functions normally. When the agent identifier of the target agent is configured to be revoked, this method instructs the target agent to stop outputting generated content.

[0035] The method according to this embodiment can be implemented by automatically executed code. To further improve the effect of forced execution, the code of this method can be written to the blockchain in the form of a smart contract.

[0036] The method according to this embodiment can be deployed in the following specific application scenarios: 1. Compliance certification of generative AI content.

[0037] Commercial users of generative AI content (such as news organizations, stock photo platforms, and AI-assisted news tools) often find it difficult to prove which model generated the content they use, whether authorized data was used, and who should bear the responsibility. The method according to this embodiment can determine a causal identification chain for each instance of generated content, used for verification and tracing.

[0038] 2. Governance of internal AI systems within enterprises.

[0039] Enterprises (such as banks, insurance companies, and human resource automation platforms) often find it difficult to track which model made which decision or which agent performed an action within an AI system. The method described in this embodiment allows for the tracing of each decision output, the selective revocation of agents, and the isolation of the risk of decision outputs into specific models.

[0040] 3. Management of multi-agent systems.

[0041] In fields involving multi-agent systems (such as AI workflow automation, autonomous trading robots, and robot control systems), there is often a problem of individual agents acting independently and a lack of accountability. According to the method of this embodiment, each agent corresponds to a unique agent identifier, enabling the tracking of the causal identification chain of decisions and the allocation of responsibility.

[0042] 4. Copyright and dataset licensing.

[0043] Owners of training datasets often lack the technical means to enforce authorization of their datasets. According to the method of this invention, the scope of authorization for the training dataset is defined by data usage constraint information in the dataset identifier. When the data usage constraint information is incompatible with the model, this method can prevent the model from using the training dataset.

[0044] 5. Regulation of AI-generated content.

[0045] Content managers (such as AI registries and national AI security agencies) typically lack the capacity to oversee AI-generated content and effectively audit AI systems. According to the method described in this invention, content managers can perform identifier checks as needed, without accessing the original data, and provide cryptographic proofs of compliance.

[0046] The method according to embodiments of the present invention overcomes the problems of fragmentation, opacity, and lack of traceability in existing AI governance mechanisms. It automatically executes throughout the entire lifecycle of AI-generated content, dynamically and comprehensively statistically analyzes the information and information subjects involved in the lifecycle across multiple levels, and performs compatibility analysis, status recording, verification, and accountability based on the statistical identifiers, ensuring that the generated content output by the intelligent agent each time can be reliably traced back to the training dataset, model instance, and intelligent agent.

[0047] Figure 4 A schematic block diagram of a content generation management system for intelligent agents according to an embodiment of the present invention is shown. Figure 4As shown, the management system 400 for generating content for intelligent agents includes an information digest determination module 401, an intelligent agent identifier determination module 402, a model identifier determination module 403, and a causal identification chain determination module 404. The information digest determination module 401 is configured to obtain an information digest corresponding to the generation condition information input from the target intelligent agent. The intelligent agent identifier determination module 402 is configured to obtain the intelligent agent identifier of the target intelligent agent. The model identifier determination module 403 is configured to obtain the model identifier of each of the one or more target models in response to the target intelligent agent determining one or more target models based on the generation condition information. The causal identification chain determination module 404 is configured to determine the causal identification chain of the generated content based on the information digest, the intelligent agent identifier, and the model identifier of each of the one or more target models in response to the target intelligent agent outputting generated content based on one or more target models, and store the causal identification chain in a distributed ledger or trusted evidence storage system.

[0048] In some embodiments, the causal identification chain determination module 404 may be further configured to: determine the state digest of the causal identification chain and store the state digest in a distributed ledger or trusted evidence storage system.

[0049] In some embodiments, the causal identification chain determination module 404 may be further configured to: determine the causal identification chain based on the output time, information digest, agent identifier, and model identifier of each target model in one or more target models of the target agent's output generated content.

[0050] In some embodiments, the agent identifier may include: the scope of action authority of the target agent, and the responsibility attribution identifier of the owner of the target agent.

[0051] In some embodiments, the model identifier of any target model may include: a dataset identifier for each of one or more training datasets used to train the target model; at least one of resource, system, and process feature information used to train the target model; the architecture of the target model; and descriptive information for describing the target model.

[0052] In some embodiments, the model identifier of any target model may further include: performance metrics and policy compliance assessment results added after periodic testing of the target model; and abnormal behavior metrics added after periodic testing of the target model and detection of anomalies in the target model.

[0053] In some embodiments, the dataset identifier for any training dataset may include: data source information, data ownership information, and data usage constraint information.

[0054] In some embodiments, the system may further include a compatibility detection module configured to: analyze whether the target agent is compatible with one or more target models based on the agent identifier and the model identifier of each of the one or more target models; and instruct the target agent to stop outputting generated content if the target agent is incompatible with at least one of the one or more target models.

[0055] In some embodiments, the system may further include a verification module configured to verify the causal identification chain using hash proofs and ledger checks.

[0056] In some embodiments, the verification module is further configured to verify the causal identification chain using hash proofs, ledger checks, and zero-knowledge proofs.

[0057] Figure 5 A schematic diagram of a computer system is shown, which can implement the method and apparatus for managing generated content for intelligent agents according to embodiments of the present invention. It should be understood that... Figure 5 The computer system 500 shown is merely an example and should not be construed as limiting the functionality and scope of use of the method and apparatus for managing generated content for intelligent agents according to embodiments of the present invention.

[0058] like Figure 5 As shown, the computer system 500 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage device 508 into a random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the computer system 500. The processing device 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0059] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, cameras, accelerometers, gyroscopes, sensors, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, motors, electronic speed controllers, etc.; storage devices 508 including, for example, flash cards; and communication devices 509. Communication device 509 allows computer system 500 to communicate wirelessly or wiredly with other devices to exchange data. Although... Figure 5 A computer system 500 with various devices is shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have instead. Figure 5 Each box shown can represent a device or multiple devices as needed.

[0060] In particular, according to some embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer programs. For example, a computer-readable medium is provided having a computer program stored thereon, the computer program comprising methods for executing... Figure 2 The program code shown is for a method of managing generated content for intelligent agents. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by the processing device 501, the functional units described above in the device for managing generated content for intelligent agents according to an embodiment of the present invention are implemented.

[0061] It should be noted that the computer-readable medium according to embodiments of the present invention may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. A computer-readable storage medium according to embodiments of the present invention may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. Additionally, a computer-readable signal medium according to embodiments of the present invention may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (Radio Frequency), etc., or any suitable combination thereof.

[0062] Computer program code for performing operations according to embodiments of the present invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0063] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0064] This invention can be implemented in other specific forms without departing from its spirit and essential characteristics. For example, the algorithm described in a particular embodiment can be modified without departing from the basic spirit of the invention. Therefore, the embodiments are to be regarded as exemplary rather than limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and all changes falling within the meaning and scope of the claims and their equivalents are thus included within the scope of the invention.

Claims

1. A method for managing generated content for intelligent agents, comprising: In response to the input generation condition information of the target intelligent agent, obtain the information digest corresponding to the generation condition information; Obtain the agent identifier of the target agent; In response to the target agent determining one or more target models based on the generation condition information, the current model identifier of each of the one or more target models is obtained; as well as In response to the target agent outputting generated content based on the one or more target models, a causal identification chain of the generated content is determined based on the information digest, the agent identifier, and the model identifier of each target model in the one or more target models, and the causal identification chain is stored in a distributed ledger or trusted evidence storage system.

2. The method according to claim 1, wherein, Storing the causal identification chain in a distributed ledger or trusted evidence storage system includes: storing a state summary of the causal identification chain in the distributed ledger or trusted evidence storage system.

3. The method according to claim 1, wherein, Determining the causal identification chain includes: The causal identification chain is determined based on the output time of the generated content output by the target agent, the information digest, the agent identifier, and the model identifier of each target model in the one or more target models.

4. The method according to claim 1, wherein, The agent identifier includes: the action authority scope of the target agent, and the responsibility attribution identifier of the owner of the target agent.

5. The method according to claim 1, wherein, The model identifier for any target model includes: A dataset identifier for each of one or more training datasets used to train the target model; At least one of the resource, method, and process feature information used to train the target model; The architecture of the target model; and Descriptive information used to describe the target model.

6. The method according to claim 5, wherein, The model identifier for any target model also includes: The performance metrics and policy compliance assessment results added after the target model is periodically tested; An abnormal behavior indicator added after periodically testing the target model and detecting anomalies in the target model.

7. The method according to claim 5, wherein, The dataset identifier for any training dataset includes: data source information, data ownership information, and data usage constraints information.

8. The method according to claim 1, wherein, Also includes: Based on the agent identifier and the model identifier of each of the one or more target models, analyze whether the target agent is compatible with the one or more target models; If the target agent is incompatible with at least one of the one or more target models, the target agent is instructed to stop outputting the generated content.

9. The method according to claim 1, wherein, Also includes: The causal identification chain is verified using hash proofs and ledger checks.

10. The method according to claim 9, wherein, Verifying the causal identification chain includes using hash proofs, ledger checks, and zero-knowledge proofs to verify the causal identification chain.

11. A management system for generating content for intelligent agents, comprising: The information digest determination module is configured to obtain an information digest corresponding to the generation condition information in response to the input generation condition information of the target agent. The agent identifier determination module is configured to obtain the agent identifier of the target agent. The model identifier determination module is configured to, in response to the target agent determining one or more target models based on the generation condition information, obtain the model identifier of each of the one or more target models; as well as The causal identification chain determination module is configured to, in response to the target agent outputting generated content based on the one or more target models, determine the causal identification chain of the generated content based on the information digest, the agent identifier, and the model identifier of each of the one or more target models, and store the causal identification chain in a distributed ledger or trusted evidence storage system.

12. The system according to claim 11, wherein, The causal identification chain determination module is further configured to: determine the state digest of the causal identification chain and store the state digest in a distributed ledger or trusted evidence storage system.

13. The system according to claim 11, wherein, The causal identification chain determination module is further configured to: determine the causal identification chain based on the output time of the generated content output by the target agent, the information digest, the agent identifier, and the model identifier of each target model in the one or more target models.

14. The system according to claim 11, wherein, The agent identifier includes: the action authority scope of the target agent, and the responsibility attribution identifier of the owner of the target agent.

15. The system according to claim 11, wherein, The model identifier for any target model includes: A dataset identifier for each of one or more training datasets used to train the target model; At least one of the resource, system, and process feature information used to train the target model; The architecture of the target model; and Descriptive information used to describe the target model.

16. The system according to claim 15, wherein, The model identifier for any target model also includes: The performance metrics and policy compliance assessment results added after the target model is periodically tested; An abnormal behavior indicator added after periodically testing the target model and detecting anomalies in the target model.

17. The system according to claim 15, wherein, The dataset identifier for any training dataset includes: data source information, data ownership information, and data usage constraints information.

18. The system according to claim 11, wherein, It also includes a compatibility detection module, which is configured as follows: Based on the agent identifier and the model identifier of each of the one or more target models, analyze whether the target agent is compatible with the one or more target models; If the target agent is incompatible with at least one of the one or more target models, the target agent is instructed to stop outputting the generated content.

19. The system according to claim 11, wherein, It also includes a verification module configured to verify the causal identification chain using hash proofs and ledger checks.

20. The system according to claim 19, wherein, The verification module is further configured to verify the causal identification chain using hash proofs, ledger checks, and zero-knowledge proofs.

21. An apparatus for managing generated content of an intelligent agent, comprising: processor; as well as A memory having stored computer-executable instructions thereon, wherein, when executed by the processor, the computer-executable instructions cause the processor to perform the method for managing generated content for an intelligent agent as described in any one of claims 1 to 10.

22. A computer-readable storage medium having stored thereon computer-executable instructions, wherein, When executed by a processor, the computer-executable instructions cause the processor to perform the method for managing generated content for an intelligent agent as described in any one of claims 1 to 10.

23. A computer program product comprising computer-executable instructions, wherein, When executed by a processor, the computer-executable instructions cause the processor to perform the method for managing generated content for an intelligent agent as described in any one of claims 1 to 10.

Images