For the first time, guidelines are drafted to provide industry-specific rules for companies and content creators. Videos must contain an "explicit" label. Experts foresee difficulties in application.

 

Beijing (AsiaNews) – China is planning new legislation to ensure that contents generated by artificial intelligence (AI) are clearly labelled; its goal is to stem the tide of fraud related to the use of this new technology.

Published on 14 September, the draft guidelines, which are open to public comments for a month, would require all AI-generated images, videos, and audios to be clearly labelled with an electronic watermark and embedded metadata.

This is the first time that China’s Cyberspace Administration proposes specific rules to define and oversee AI-generated content.

Under the guidelines, generated videos must contain an "explicit" label displayed at the start of the clip, which must be visible at all times in a corner of the screen.

One major recommendation calls for flashing up a label at “appropriate” moments during the video.

Metadata recording the file’s source and copyright information — referred to in the guidelines as an “implicit” label — must also be logged at the time of creation.

The rules would apply not only to AI companies, but also to individual content creators, online platforms, app stores, and any other content distributors.

Going one step further, the guidelines would require content distribution platforms to label files suspected of being AI-generated in case metadata are missing.

Finally, app stores must ensure that content providers correctly label AI-generated content.

These proposals are the latest effort by Chinese authorities to combat the rise in fraud cases linked to the new technology.

According to Chinese start-up RealAI, more than 185 million yuan (US$ 26 million) were stolen in China in the first five months of 2024 using AI, compared to just 16.7 million yuan in the whole of 2023.

Deepfake technology, text-to-speech, and AI chatbots are listed as "typical high-risk application scenarios" in the guidelines, highlighting their potential use to defraud users.

Still, analysts and experts point out that the new directives may not be so easy to apply in practice.

Ma Ce, a lawyer with the Zhejiang Kinding Law Firm and a specialist in Internet law, is one such critical voices. In his view, the guidelines lack a clear definition of what constitutes "AI-generated content requiring watermarking".

The obligation for creators and online platforms to properly add metadata to generated files will pose technical challenges, he added.

The guidelines also do not specify penalties for creators and distributors of content who violate the rules.

China is not the only country that is trying to ensure that AI-generated content is clearly labelled. The European Union, the United States, Singapore, and Canada are moving in the same direction.

Several questions remain, though, about how the rules requiring watermarking online content can be implemented.

• china
• artificial intelligence
• content generation
• watermarking
• guidelines
• metadata
• copyright