Friend
Professional
- Messages
- 2,653
- Reaction score
- 850
- Points
- 113
AI content creators face a difficult choice – labeling or a fine.
The Internet Affairs Administration of China (CAC) has submitted a draft of new rules aimed at regulating the labeling of AI content. The document was developed in order to protect national security, public interests, as well as the rights and legitimate interests of citizens, organizations and companies.
The main task of the new document is to create uniform standards for labeling content that is generated or synthesized using AI. The draft rules define two types of marking:
The draft rules specifically require that when a user downloads, copies or exports AI content, all specified labels and markings must be retained. Service providers are required to ensure that these labels are in place at all stages of the content experience.
It is important that the rules also apply to online platforms that distribute content on the Internet. Services should check for labeled metadata and display the relevant information to users. For example, if the platform detects that content contains hidden markings, users will be provided with an explicit notification that they are looking at AI-generated material. In the event that there is no labeling, but the user or service provider claims that the content was generated by AI, the platform must also inform the audience about this.
The draft rules introduce a number of requirements for services that distribute content using AI. For example, companies must implement tools in their systems to ensure that the labels are displayed correctly on all materials. For example, for text and images, labels should be placed at the beginning or end of the file, and in the case of videos, on the initial frames.
App distribution platforms are also required to verify that developers have such tools to ensure content labeling before allowing software products to be placed on platforms.
The rules also require companies providing AI services to include information about labeling methods and standards in user agreements. Users should be aware of how AI-generated content is labeled and the obligations they have when using such content.
In addition, users who intend to distribute content created using AI are required to independently indicate its origin. Any attempt to remove, alter or conceal the marking is prohibited and may result in legal consequences.
The final section of the draft rules establishes measures of liability for violation of these requirements. Companies that do not comply with the labeling rules will be held administratively liable.
At the moment, the project is at the stage of public discussion, which will last until October 14, 2024, after which the document will be finalized taking into account the comments received. The new rules will come into force after the approval and publication of the final version of the document.
Source
The Internet Affairs Administration of China (CAC) has submitted a draft of new rules aimed at regulating the labeling of AI content. The document was developed in order to protect national security, public interests, as well as the rights and legitimate interests of citizens, organizations and companies.
The main task of the new document is to create uniform standards for labeling content that is generated or synthesized using AI. The draft rules define two types of marking:
- Explicit labeling is those visual or auditory elements that users can easily notice in AI-generated materials. These can be text or graphic icons, as well as beeps that indicate that the content has been generated using AI.
- Covert marking, in turn, is embedded in the file's metadata and is designed to be more difficult to detect. The mark includes data such as content attributes, the name or code of the service provider, and the file number.
The draft rules specifically require that when a user downloads, copies or exports AI content, all specified labels and markings must be retained. Service providers are required to ensure that these labels are in place at all stages of the content experience.
It is important that the rules also apply to online platforms that distribute content on the Internet. Services should check for labeled metadata and display the relevant information to users. For example, if the platform detects that content contains hidden markings, users will be provided with an explicit notification that they are looking at AI-generated material. In the event that there is no labeling, but the user or service provider claims that the content was generated by AI, the platform must also inform the audience about this.
The draft rules introduce a number of requirements for services that distribute content using AI. For example, companies must implement tools in their systems to ensure that the labels are displayed correctly on all materials. For example, for text and images, labels should be placed at the beginning or end of the file, and in the case of videos, on the initial frames.
App distribution platforms are also required to verify that developers have such tools to ensure content labeling before allowing software products to be placed on platforms.
The rules also require companies providing AI services to include information about labeling methods and standards in user agreements. Users should be aware of how AI-generated content is labeled and the obligations they have when using such content.
In addition, users who intend to distribute content created using AI are required to independently indicate its origin. Any attempt to remove, alter or conceal the marking is prohibited and may result in legal consequences.
The final section of the draft rules establishes measures of liability for violation of these requirements. Companies that do not comply with the labeling rules will be held administratively liable.
At the moment, the project is at the stage of public discussion, which will last until October 14, 2024, after which the document will be finalized taking into account the comments received. The new rules will come into force after the approval and publication of the final version of the document.
Source
