China Proposes Mandatory Labeling for AI-Generated Content Online

China proposes mandatory red flags placed on all AI-generated content posted online

The notoriously strict Cyberspace Administration of China wants to regulate generative AI. On Saturday, the Cyberspace Administration of China announced a likely-to-be-enforced draft plan to enforce mandatory red-flagging of all generative AI content posted online in China [h/t The Register]. Flagging methods include audible disclaimers in audio files, visible watermarks in images, and notices spread throughout the video. Users who don't flag their AI content may still have it flagged anyway, based on metadata or its absence. The primary purpose of this being a "draft" plan until October, despite the fact that China will enforce this as it pleases, seems to be to give platforms a chance to prepare since the public comments will likely not factor into the final version of this plan. As The Register points out in its coverage, widespread regulation like this (particularly in censorship) is typical of how China treats the Internet -- but that doesn't necessarily make this a bad idea. However, generative AI and its loudest proponents are openly seeking to replace skilled labor and artists however they possibly can, and mandatory disclaimers that what you're posting is, in fact, AI-generated and not the fruits of your labor might be...a good thing? Several ongoing lawsuits relate to generative AI here in the U.S. and elsewhere. Public comments from the likes of OpenAI's Sam Whitman openly state that it's "impossible" to create AI tools without using copyrighted material -- something we in the universe where we still have to work for our money recognize as copyright infringement. At this point, the cat's out of the bag, so the technology will almost certainly stay -- but in its current and near-unregulated form, where it's almost purely destructive to the environment and skilled workers and shows unprecedented power demand? That seems unlikely. At least in China, passing off generative AI's work as your own will be near-impossible before the end of the year.

China wants red flags on all AI-generated content

Visible and audible warnings, plus metadata, with absence of info considered suspicious China's internet regulator on Saturday proposed a strict regime that will, if adopted, require digital platforms to label content created by artificial intelligence. The Cyberspace Administration of China announced its draft plan, which will require platforms and online service providers to label all AI-generated material with a visible logo and with metadata embedded in relevant files. The draft proposes that logos appear in several locations in a text, image, video, or audio file. In audio files, Beijing wants a voice prompt to inform listeners about AI-generated content at the start and end of a file - and, as appropriate, mid-file too. Software that plays audio files will also need to inform netizens when they tune in to AI content. Video players can get away with just posting notices about the content at the start, end, and relevant moments during a clip. Netizens who post AI-generated content will be required to label it as such. If they use generation tools provided by a platform, they'll be required to identify themselves - and a log of their activities will be retained for six months. Some labels denoting AI-made content will be applied dynamically, based on metadata embedded in AI-generated content. If metadata is absent from a file, service providers will be required to analyze content. If an org suspects that a file was made by AI, it will need to label it as such. The draft code calls for China's internet platforms to share the smarts they develop as they work to detect AI content, so that all can improve their efforts. Failing to do any or the above will mean trouble for platform operators and individuals. The draft is just that, and comments on it are open until October. However Chinese regulators aren't famed for making big changes after drafts like this appear - this draft is likely the regime Beijing wants to see implemented. It's pretty much business as usual: China always wants its tech giants to police what comes online, and to tie user-generated content to an identifiable individual. Doing so means China can enact its vision of sovereign control over its local internet - which translates into criticism of the Communist Party or government being extremely unlikely to make it online, or persist for long if its does. ®

China to Require All AI-Generated Content to Be Labeled Clearly

The Chinese internet authority has proposed strict regulations for AI images requiring them to be clearly labeled and have embedded metadata. The Cyberspace Administration of China unveiled its draft plan over the weekend which recommends all AI-generated content, including pictures, video, and audio, to have clear -- even multiple -- labels. AI videos, for example, must post notices at the start and end of the video as well as during "appropriate" times during the clip. If adopted, the rules will apply to AI companies, educational and scientific institutions, cultural institutions, and everyday internet users. Users of generative AI tools will be required to identify themselves and a log of their usage will be retained for six months. As well as visible watermarks, companies and people will need to embed metadata into AI content. If the metadata is absent from a file, tech firms will be required to analyze the content and if it is suspected of being AI-generated then the company will be compelled to label it as such. Those who don't adhere to these rules will find themselves in trouble. The Register notes that while the proposed rules are still at the draft stage and public consultation will continue until October, Chinese regulators aren't known for making changes once statutes like this are posted. The tech publication says that these are the regulations Beijing likely wants to see implemented for AI. Zhang Linghan, a professor at the China University of Political Science and Law, tells the People's Daily, the official newspaper of the Central Committee of the Chinese Communist Party, that "the labeling system will enhance governance capabilities for AI-generated content, reduce the production of misinformation, prevent the contamination of future training data sets, and foster a healthy cyber information ecosystem". "This initiative marks a new phase of stringent regulation and high compliance standards in China's generative AI services sector," a Beijing-based lawyer who specializes in intellectual property transactions tells The South Morning China Post. AI regulation is a hot topic around the world in 2024; Adobe, OpenAI, and Microsoft have backed a bill in California that requires tech companies to clearly label AI-generated content. Last week, a host of countries including the United States and the United Kingdom signed the "first-ever international legally binding treaty" aimed at ensuring AI systems are "fully consistent with human rights, democracy and the rule of law."