China’s draft rules call for clear identification of AI-generated content

The Cyberspace Administration of China is considering labeling all content created by artificial intelligence as part of a new draft rule. According to reports, the regulator, which is the Chinese internet authority, has already prepared a draft plan asking all digital platforms and online service providers to add the AI mark to AI-generated content.

In an announcement on September 14, the regulator said this move would standardize the content identification process for AI-generated content and protect national security and the interests of individuals and organizations using the platforms.

The regulator released a draft law open for public comments until October 14. According to the announcement, the public can send comments via email to biaoshi@cac.gov.cn or by letter to the Network Management Technology Bureau of the State Internet Information Office.

While public comments could impact the final draft of the rules, experts believe the proposed rules would likely be passed in their current form with minimal changes as the Chinese authorities are not known for making drastic changes to draft rules based on comments.

Draft rules require adding logos and voice prompts to AI content

The draft rule proposes that any text, video, or image file created by AI should include a logo to signify that it is AI-generated, effectively notifying anyone who comes across such content. To ensure that users notice the label, the rule wants it to be shown at the start, end, and other relevant moments within a video.

For audio files, creators must add voice prompts that inform users when they are about to listen to AI-generated content, with such prompts notifying them of various parts of the audio content. All software that plays audio content would also be required to have prompts that tell users when they start listening to AI content.

Article 4 of the proposed rules reads:

“If the generation synthesis service provided by the service provider falls under the first paragraph of Article 17 of the Regulations on the Management of Deep Synthesis of Internet Information Services, an explicit logo shall be added to the generated synthesis content in accordance with the following requirements.”

Meanwhile, the draft rule also requires metadata embedded into a file to reflect that it is AI-generated, making it possible to label such content. However, where there is no metadata, it would be up to the service providers to analyze the content and label it as such if it is suspected to be AI-generated.

The rules also apply to Chinese internet users. Individuals who post AI-generated content have to label it. When they use AI tools on other platforms, the proposed law requires them to provide their personal information, and the AI-generation platform will keep a log of their activities for six months.

China requires internet platforms to share technology

Meanwhile, the draft plan also includes requirements that the social media platform must share its technology for detecting AI content with others, a move designed to standardize the process and improve the ability to detect AI content.

The new rules do not come as a surprise, given how content moderation has always been a crucial part of internet use in China. However, the focus on AI content appears to be more relevant in the face of deepfakes and other AI-generated content being used by many to commit fraud and other illegalities.

Several popular social media platforms outside of China, including Facebook, Instagram, and TikTok, already label AI content as part of their content moderation. However, their detection tools have not been perfect, as multiple hyper-realistic content still escape the AI tags on these platforms