Artificial intelligence content traceability and dynamic supervision method, system and related equipment

Abstract

The invention relates to the technical field of artificial intelligence, and discloses an artificial intelligence content traceability and dynamic supervision method and system and related equipment, and the method comprises the steps: obtaining artificial intelligence content; comparing the artificial intelligence content with violation characteristics in a violation library, and determining whether the artificial intelligence content is violated; if the artificial intelligence content is illegal, analyzing the artificial intelligence content, and determining an encryption identifier; and tracing the artificial intelligence content based on the encrypted identifier. According to the method, the artificial intelligence content is compared with the violation characteristics, dynamic supervision of the artificial intelligence content is realized, and in addition, the artificial intelligence content is traced by extracting the encryption identifier in the artificial intelligence content.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of artificial intelligence, in particular to an artificial intelligence content tracing and dynamic supervision method and system and related equipment. BACKGROUND

[0002] With the continuous development of artificial intelligence, many contents are no longer generated by artificial, but replaced by generative AI technology. For example, text-to-video technology is an important branch of artificial intelligence generated content (AIGC) field, which is to convert text description into dynamic video through natural language processing (NLP) and computer vision (CV) technology. For another example, virtual human technology creates digital images with human appearance, behavior and even thought through computer graphics (CG), motion capture and artificial intelligence technology.

[0003] However, with the popularity of generative AI, the risk of deep forgery and false information has increased dramatically, and the generated content cannot be effectively supervised and traced. SUMMARY

[0004] In order to overcome the problem that the risk of deep forgery and false information has increased dramatically in generative AI technology, and the generated content cannot be effectively supervised and traced, the present disclosure provides an artificial intelligence content tracing and dynamic supervision method and system and related equipment.

[0005] In a first aspect, in order to solve the above technical problems, the present disclosure provides an artificial intelligence content tracing and dynamic supervision method, comprising: obtaining artificial intelligence content; wherein the artificial intelligence content is a video generated based on text-to-video technology or a virtual human generated based on virtual human generation technology; comparing the artificial intelligence content with the illegal features in the illegal library to determine whether the artificial intelligence content is illegal; if the artificial intelligence content is illegal, analyzing the artificial intelligence content to determine the encryption identifier; wherein the encryption identifier is an encryption identifier embedded in the process of generating the artificial intelligence content; tracing the artificial intelligence content based on the encryption identifier.

[0006] Further, comparing the artificial intelligence content with the illegal features in the illegal library to determine whether the artificial intelligence content is illegal, comprising: extracting feature data in the artificial intelligence content; wherein the feature data includes feature data corresponding to text, voice and action in the artificial intelligence content; matching the feature data with the illegal features in the illegal library; if the feature data matches any illegal feature successfully, it is judged that the artificial intelligence content is illegal; If the feature data fails to match all the violation features, it is determined that the artificial intelligence content is not in violation.

[0007] Further, the artificial intelligence content is compared with the violation features in the violation library to determine whether the artificial intelligence content is in violation, which further includes: If the artificial intelligence content is in violation, a target violation feature is obtained; wherein the target violation feature is a violation feature that successfully matches the feature data; The target violation feature is classified to determine the violation type; The artificial intelligence content is risk-managed based on the violation type.

[0008] Further, the artificial intelligence content is risk-managed based on the violation type, which includes: Based on the violation type, a risk level is determined; If the risk level is a first-level risk, a risk label is added to the artificial intelligence content and the flow is limited; If the risk level is a second-level risk, the artificial intelligence content is forced to be taken down.

[0009] Further, if the artificial intelligence content is a video, the artificial intelligence content is parsed to determine an encryption identifier, which includes: The video is parsed to determine a frame sequence; A model fingerprint is decoded from the frame sequence; The encryption identifier includes the model fingerprint.

[0010] Further, if the artificial intelligence content is a virtual human, the artificial intelligence content is parsed to determine an encryption identifier, which includes: The virtual human is parsed to determine an action skeleton data stream; A watermark is parsed from the action skeleton data stream; The encryption identifier includes the watermark.

[0011] Further, the artificial intelligence content is traced based on the encryption identifier, which includes: Based on the model fingerprint or the watermark, developer information is determined; The developer information is pushed to a regulatory agency.

[0012] In a second aspect, the present disclosure provides an artificial intelligence content traceability and dynamic supervision system, which includes: An artificial intelligence content acquisition module is configured to acquire artificial intelligence content; wherein the artificial intelligence content is a video generated based on a text-to-video technology or a virtual human generated based on a virtual human generation technology; A violation judgment module is configured to compare the artificial intelligence content with violation features in a violation library to determine whether the artificial intelligence content is in violation; The encryption identifier determination module is used to parse the AI ​​content and determine the encryption identifier if the AI ​​content violates regulations; wherein, the encryption identifier is an encryption identifier embedded in the AI ​​content during the generation process. The traceability module is used to trace the source of artificial intelligence content based on encrypted identifiers.

[0013] Thirdly, this disclosure provides a computing device, including a memory, a processor, and a program stored in the memory and running on the processor, wherein the processor executes the program to implement the artificial intelligence content tracing and dynamic monitoring method described above.

[0014] Fourthly, this disclosure provides a computer-readable storage medium storing instructions that, when executed on a terminal device, cause the terminal device to perform the aforementioned artificial intelligence content tracing and dynamic monitoring method.

[0015] The beneficial effects of this disclosure are: acquiring artificial intelligence content and comparing it with violation characteristics to determine whether the artificial intelligence content is in violation, thereby achieving dynamic supervision of artificial intelligence content; in addition, when artificial intelligence content is in violation, the encrypted identifier in the artificial intelligence content is extracted to trace the source of the artificial intelligence content, establish a sound source tracing system, and effectively hold violators accountable. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the present disclosure will be further described below in conjunction with the accompanying drawings and embodiments.

[0017] Figure 1 This is a flowchart illustrating the artificial intelligence content tracing and dynamic monitoring method according to an embodiment of this disclosure; Figure 2 This is a schematic diagram of the structure of the artificial intelligence content tracing and dynamic monitoring system according to an embodiment of this disclosure; Figure 3 This is a schematic diagram of the structure of a computing device according to an embodiment of the present disclosure. Detailed Implementation

[0018] The following embodiments are further explanations and supplements to this disclosure and do not constitute any limitation on this disclosure.

[0019] The following describes, with reference to the accompanying drawings, the artificial intelligence content tracing and dynamic monitoring method, system and related equipment of this disclosure.

[0020] like Figure 1 As shown, this disclosure provides a method for tracing and dynamically monitoring artificial intelligence-based content, including: S1, obtain artificial intelligence content; wherein the artificial intelligence content is a video generated based on a text-to-video technology or a virtual human generated based on a virtual human generation technology.

[0021] S2, compare the artificial intelligence content with the violation features in the violation library to determine whether the artificial intelligence content is in violation.

[0022] S3, if the artificial intelligence content is in violation, analyze the artificial intelligence content to determine an encryption identifier; wherein the encryption identifier is an encryption identifier embedded in the artificial intelligence content during generation.

[0023] S4, trace the artificial intelligence content based on the encryption identifier.

[0024] In this embodiment, the artificial intelligence content is obtained and compared with the violation features to determine whether the artificial intelligence content is in violation, thereby achieving dynamic supervision of the artificial intelligence content. In addition, when the artificial intelligence content is in violation, the encryption identifier in the artificial intelligence content is extracted to trace the artificial intelligence content, establish a perfect traceability system, and effectively hold the illegal personnel accountable.

[0025] In this embodiment, the creator adds an encryption identifier to the generated artificial intelligence content when creating to ensure the originality and non-tamperability of the generated video and virtual human, and to ensure that the creator can provide effective evidence when seeking rights protection. In this embodiment, the encryption identifier can be used to effectively trace the artificial intelligence content, thereby comprehensively supervising the generated artificial intelligence content.

[0026] Optionally, the artificial intelligence content is compared with the violation features in the violation library to determine whether the artificial intelligence content is in violation, including: extracting feature data in the artificial intelligence content; wherein the feature data includes feature data corresponding to text, voice, and actions in the artificial intelligence content; matching the feature data with the violation features in the violation library; if the feature data matches any violation feature successfully, it is determined that the artificial intelligence content is in violation; if the feature data fails to match all violation features, it is determined that the artificial intelligence content is not in violation.

[0027] In this embodiment, the artificial intelligence content mainly includes characters, voices, and actions contained in the video, as well as characters, voices, and actions exhibited by the virtual human. Therefore, the feature data can be extracted through feature extraction technology, for example: When a character is included in a video or virtual human, a target detection and tracking algorithm can be used, such as using a YOLO, Faster R-CNN, etc. model to detect a human body, thereby extracting a target object, and then detecting the target object through human key point detection (such as OpenPose, MediaPipe) to obtain joint coordinates, and finally identifying the joint coordinates through time sequence action recognition (such as 3D-CNN, I3D, SlowFast) to analyze the action corresponding to the feature data of the character in the video (such as waving hands, falling down, etc.).

[0028] When a voice is included in a video or virtual human, a voice-to-text model (such as Wav2Vec2.0, Whisper, etc.) can be used to directly extract the feature data corresponding to the voice.

[0029] When text is included in a video or virtual human, the text information (such as FFmpeg tool) can be directly extracted as feature data.

[0030] In this embodiment, the violation library includes various violation features, such as violent actions, indecent actions, etc., and violation voices, such as sensitive words, hate speech, etc. The feature data is matched with all the violation features. If the feature data matches any violation feature, it indicates that the artificial intelligence content has violation content. If the feature data does not match any violation feature, it indicates that the artificial intelligence content does not have violation content.

[0031] Optionally, the artificial intelligence content is compared with the violation features in the violation library to determine whether the artificial intelligence content is in violation, which also includes: If the artificial intelligence content is in violation, the target violation feature is obtained; wherein the target violation feature is the violation feature that matches the feature data successfully; The target violation feature is classified to determine the violation type; The artificial intelligence content is managed based on the violation type.

[0032] In this embodiment, different violation features correspond to different violation types, which are illustrated as follows: (1) The violation content corresponding to the voice includes sensitive words, but different degrees of sensitive words correspond to different violation types. For example, if the sensitive words include some discriminatory language or exaggerated effect words, the violation type can be defined as inappropriate language, and religious, pornographic, etc. words can be defined as strictly prohibited.

[0033] (2) The violation content corresponding to the action includes violent behavior, but different degrees of violent behavior correspond to different violation content. For example, if the violent behavior is a slight push, the violation type can be defined as slight violence, and if the violent behavior is mutual fighting, the violation type can be defined as serious violence.

[0034] In this embodiment, a logistic regression model, a decision tree model, and a k-nearest neighbor model (KNN) can be used to classify the target violation features.

[0035] Optionally, the risk management of the artificial intelligence content based on the violation type includes: determining a risk level based on the violation type; if the risk level is a first-level risk, adding a risk label to the artificial intelligence content and limiting the flow; if the risk level is a second-level risk, forcibly removing the artificial intelligence content.

[0036] In this embodiment, different violation types are processed by grading, making the dynamic supervision of the artificial intelligence content more reasonable, and avoiding excessive supervision.

[0037] Optionally, if the artificial intelligence content is a video, the artificial intelligence content is parsed to determine an encryption identifier, including: parsing the video to determine a frame sequence; decoding a model fingerprint from the frame sequence; wherein the encryption identifier includes the model fingerprint.

[0038] In this embodiment, the model fingerprint (Model Fingerprint) is a digital marker used to uniquely identify a generated model. Its essence is to embed extractable hidden features in generated content through algorithms, for example, in the video generation process, by modifying the noise distribution in the latent space of the video diffusion model (such as Stable Video Diffusion) to embed the fingerprint.

[0039] In this embodiment, when extracting the model fingerprint, the frame sequence is subjected to DCT / DWT transformation, the energy anomaly (such as the model fingerprint) of a specific frequency band is detected, the features at the energy anomaly are extracted, and then the extracted features are compared with the known model fingerprint library (such as cosine similarity, Euclidean distance), or the extracted features are directly classified using a classification model (such as SVM, MLP), to finally obtain the model fingerprint.

[0040] Optionally, if the artificial intelligence content is a virtual human, the artificial intelligence content is parsed to determine an encryption identifier, including: parsing the virtual human to determine a motion skeleton data stream; parsing a watermark from the motion skeleton data stream; wherein the encryption identifier includes the watermark.

[0041] The motion skeletal data stream in the embodiment is the core data format for driving the motion of the virtual human. In essence, it is a time-sequenced skeletal joint transformation parameter, which functions like the "nervous system" of a real human body, and realizes natural motion performance of the virtual human through accurate control of skeletal motion.

[0042] In the embodiment, the creator embeds the watermark in the motion skeletal data stream. When supervising the virtual human, the motion skeletal data stream needs to be subjected to DCT / DWT transformation, the frequency band (such as the medium-high frequency coefficient) where the watermark may be embedded is located, the features at the frequency band are extracted, and then the classifier (CNN or LSTM) is used to classify the extracted features, and finally the watermark is obtained.

[0043] Optionally, the artificial intelligence content is traced based on the encrypted identifier, including: determining the developer information based on the model fingerprint or the watermark; pushing the developer information to a supervisory agency.

[0044] In the embodiment, the model fingerprint or the watermark contains a company ID or a creator ID. The developer information can be determined by extracting the model fingerprint or the watermark, and the developer information is pushed to the supervisory agency, thereby achieving the purpose of tracing.

[0045] As shown in Figure 2 The present disclosure provides an artificial intelligence content tracing and dynamic supervision system, including: an artificial intelligence content acquisition module, configured to acquire artificial intelligence content; wherein the artificial intelligence content is a video generated based on text-to-video technology, or a virtual human generated based on virtual human generation technology; a violation judgment module, configured to compare the artificial intelligence content with violation features in a violation library, and determine whether the artificial intelligence content is in violation; an encrypted identifier determination module, configured to, if the artificial intelligence content is in violation, analyze the artificial intelligence content and determine an encrypted identifier; wherein the encrypted identifier is an encrypted identifier embedded in the process of generating the artificial intelligence content; a tracing module, configured to trace the artificial intelligence content based on the encrypted identifier.

[0046] Optionally, the violation judgment module is specifically configured to: extract feature data in the artificial intelligence content; wherein the feature data includes feature data corresponding to text, voice and motion in the artificial intelligence content; match the feature data with violation features in the violation library; if the feature data matches any violation feature successfully, it is determined that the artificial intelligence content is in violation; If the feature data fails to match all the violation features, it is determined that the artificial intelligence content is not in violation.

[0047] Optionally, the violation judgment module is further configured to: If the artificial intelligence content is in violation, a target violation feature is obtained; the target violation feature is a violation feature that matches the feature data successfully. The target violation feature is classified to determine a violation type. The artificial intelligence content is subjected to risk management based on the violation type.

[0048] Optionally, the violation judgment module is specifically configured to: Based on the violation type, a risk level is determined. If the risk level is a first-level risk, a risk label is added to the artificial intelligence content and the artificial intelligence content is throttled. If the risk level is a second-level risk, the artificial intelligence content is forced to be taken down.

[0049] Optionally, the encryption identifier determination module is specifically configured to: The video is parsed to determine a frame sequence. The model fingerprint is decoded from the frame sequence. The encryption identifier includes the model fingerprint.

[0050] Optionally, the encryption identifier determination module is specifically configured to: The virtual human is parsed to determine action skeleton data flow. The watermark is parsed from the action skeleton data flow. The encryption identifier includes the watermark.

[0051] Optionally, the traceability module is specifically configured to: Based on the model fingerprint or the watermark, developer information is determined. The developer information is pushed to a regulatory agency.

[0052] A computing device of an embodiment of the present disclosure includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the artificial intelligence content traceability and dynamic supervision method described above is implemented. That is, a computing device of an embodiment of the present disclosure can include but is not limited to a processor and a memory. The memory is configured to store a computer program. The processor is configured to execute the artificial intelligence content traceability and dynamic supervision method shown in any embodiment of the present disclosure by invoking the computer program.

[0053] In one optional embodiment, a computing device is provided, as shown in Figure 3 Figure 3 ​The illustrated computing device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the computing device 4000 can also include a transceiver 4004, which can be used for data interaction, such as data transmission and / or data reception, between the computing device and other computing devices. It should be noted that the transceiver 4004 is not limited to one in actual applications, and the structure of the computing device 4000 does not constitute a limitation on the embodiments of the present disclosure.

[0054] The processor 4001 can be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It can implement or execute various exemplary logical blocks, modules and circuits described in combination with the present disclosure. The processor 4001 can also be a combination of computing functions, such as one or more microprocessor combinations, combinations of DSP and microprocessor, etc.

[0055] The bus 4002 can include a path for transmitting information between the above-mentioned components. The bus 4002 can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 4002 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one bus 4002 is represented by a thick line in the middle, but it does not mean that there is only one bus or only one type of bus.

[0056] The memory 4003 can be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.

[0057] The memory 4003 is used to store application code (computer program) for implementing the scheme of the present disclosure, and is controlled by the processor 4001 to execute. The processor 4001 is used to execute the application code stored in the memory 4003 to realize the content shown in the foregoing method embodiments.

[0058] The computing device can also be a terminal device, which can be any device that can install an application, including at least one of a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, a smart television, and a smart vehicle device.

[0059] It should be noted that, Figure 3 The computing device shown is only an example and should not limit the functions and use range of the embodiments of the present disclosure.

[0060] The computer readable storage medium of the embodiments of the present disclosure, the computer readable storage medium has a computer program stored thereon, and the computer program is executed by a processor to implement the above-mentioned artificial intelligence content traceability and dynamic supervision method.

[0061] Optionally, the computer readable storage medium can be a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0062] In an example embodiment, a computer program product or computer program is also provided, the computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. A processor of a computing device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computing device to perform the above-mentioned artificial intelligence content provenance and dynamic supervision method.

[0063] Computer program code for carrying out operations of the present disclosure can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0064] It should be understood that the flow and block diagrams in the drawings show only the architecture, functionality, and operation of possible implementations of methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flow and block diagrams can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks may

[0065] The computer readable storage medium provided by the embodiments of the present disclosure can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, the computer readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus or device.

[0066] The computer readable storage medium described above carries one or more programs, which, when executed by the computing device, cause the computing device to perform the methods shown in the embodiments described above.

[0067] The above description is merely the preferred embodiments of the present disclosure and the explanation of the technical principles applied. It should be understood by those skilled in the art that the disclosed scope of the present disclosure is not limited to the technical solutions with the specific combinations of the above technical features, and should also cover other technical solutions formed by the combinations of the above technical features or their equivalent features without departing from the technical concepts disclosed above. For example, the technical solutions formed by the mutual replacement of the above features and the technical features disclosed in the present disclosure (but not limited to) with similar functions.

[0068] It should be noted that the terms "first", "second" and the like in the specification and claims of the present application are used to distinguish similar objects, and represent no specific order or sequential order. The order of use of similar objects can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described.

[0069] Those skilled in the art know that the present disclosure can be implemented as a system, method or computer program product, therefore, the present disclosure can be specifically implemented as follows: it can be a complete hardware, a complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, which is generally referred to as "circuit", "module" or "system" herein. In addition, in some embodiments, the present disclosure can also be implemented as a computer program product in one or more computer readable media, which contains computer readable program code.

[0070] Although the embodiments of the present disclosure have been shown and described above, it is understood that the above-described embodiments are exemplary, and it is not construed that the present disclosure is limited to the above-described embodiments, and a person of ordinary skill in the art can make changes, modifications, replacements, and variations to the above-described embodiments within the scope of the present disclosure.

Claims

1. A method for tracing and dynamically monitoring content related to artificial intelligence, characterized in that, include: Acquire artificial intelligence content; wherein, the artificial intelligence content is a video generated based on text-to-video technology, or a virtual human generated based on virtual human generation technology; The AI ​​content is compared with the violation characteristics in the violation database to determine whether the AI ​​content violates regulations. If the AI ​​content violates regulations, the AI ​​content is parsed to determine an encrypted identifier; wherein the encrypted identifier is an encrypted identifier embedded in the AI ​​content during the generation process. The source of the artificial intelligence content is traced based on the encrypted identifier.

2. The method according to claim 1, characterized in that, The AI ​​content is compared with violation characteristics in a violation database to determine whether the AI ​​content violates regulations, including: Extract feature data from the AI ​​content; wherein, the feature data includes feature data corresponding to text, speech, and actions in the AI ​​content; The feature data is matched with the violation features in the violation database; If the feature data matches any of the violation features, the AI ​​content is determined to be in violation. If the feature data fails to match any of the violation features, the AI ​​content is deemed not to violate the rules.

3. The method according to claim 2, characterized in that, The process of comparing the AI ​​content with violation characteristics in a violation database to determine whether the AI ​​content violates regulations also includes: If the AI ​​content violates the rules, then the target violation feature is obtained; wherein, the target violation feature is the violation feature that successfully matches the feature data; The target violation features are classified to determine the violation type; Risk management is conducted on the AI ​​content based on the aforementioned violation type.

4. The method according to claim 3, characterized in that, The risk management of the AI ​​content based on the violation type includes: Based on the aforementioned types of violations, the risk level is determined; If the risk level is Level 1, then a risk label will be added to the AI ​​content and its reach will be limited; If the risk level is Level 2, the AI ​​content will be forcibly removed.

5. The method according to claim 1, characterized in that, If the AI ​​content is video, the step of parsing the AI ​​content and determining the encryption identifier includes: The video is analyzed to determine the frame sequence; Decode the model fingerprint from the frame sequence; The encrypted identifier includes a model fingerprint.

6. The method according to claim 5, characterized in that, If the AI ​​content is a virtual human, the step of parsing the AI ​​content and determining the encrypted identifier includes: The virtual human is analyzed to determine the motion skeleton data stream; The watermark is parsed from the motion skeleton data stream; The encrypted identifier includes the watermark.

7. The method according to claim 6, characterized in that, The process of tracing the source of the AI ​​content based on the encrypted identifier includes: Developer information is determined based on the model fingerprint or the watermark; The developer information was pushed to the regulatory authorities.

8. An AI-powered content traceability and dynamic monitoring system, characterized in that: include: An AI content acquisition module is used to acquire AI content; wherein, the AI ​​content is a video generated based on text-to-video technology, or a virtual human generated based on virtual human generation technology; The violation judgment module is used to compare the AI ​​content with the violation characteristics in the violation database to determine whether the AI ​​content violates regulations. An encryption identifier determination module is used to parse the AI ​​content and determine the encryption identifier if the AI ​​content violates regulations; wherein the encryption identifier is an encryption identifier embedded in the AI ​​content during the generation process. The tracing module is used to trace the source of the artificial intelligence content based on the encrypted identifier.

9. A computing device, comprising a memory, a processor, and a program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the artificial intelligence content tracing and dynamic supervision method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a terminal device, cause the terminal device to perform the artificial intelligence content tracing and dynamic monitoring method as described in any one of claims 1-7.

Images