China this week launched new regulations on content generated by artificial intelligence (AI). The text, which includes 24 new rules, has apparently been watered down compared to a stricter draft published at the start of 2023. Beijing’s objective is not to discourage Chinese companies from embarking on this sector, currently mainly dominated by their American counterparts.
There is growing concern around the world about generative AI, as this technology can be used for malicious purposes. Like with “deepfake” videos that make people say things they didn’t say. At the same time, Chinese companies are trying to design artificial intelligence services as efficient as ChatGPT launched by the American OpenAI, software which is currently blocked in China.
Adherence to “the fundamental values of socialism”
Generative AI must “adhere to the core values of socialism,” not threaten national security or promote terrorism, violence, or “racial hatred,” the text says. Service providers must present content generated by artificial intelligence as such, and prevent discrimination based on gender, age and ethnic group when designing algorithms. Their software must not create content containing “false and harmful information”.
Data used for training AI software must be obtained lawfully and not infringe the intellectual property rights of others.
Protected minor users
Companies that make generative AI software available to the general public must “take effective measures to prevent underage users from being excessively dependent” on these services or “becoming dependent on them”, according to rules published last July. by China’s internet regulator.
These companies must also allow users to report inappropriate content. They also have the obligation to quickly remove any illegal content. Service providers must conduct security assessments and submit reports to authorities on their algorithms. Surprise: this rule will ultimately only apply to companies whose software is considered to have an influence on “public opinion” and not to all of them, as planned in the initial project.
An offense finally less full-bodied
The new rules are technically “provisional measures” subject to pre-existing Chinese laws. These are the latest in a series of regulations targeting various aspects of artificial intelligence. “From the outset, and somewhat differently from the European Union (EU), China took a more vertical approach” to legislation, “focusing more on specific issues,” according to law firm Taylor Wessing.
The draft regulations provided for a fine of up to 100,000 yuan (12,600 euros) for violations. But under the new rules, an infraction will only be subject to a warning or suspension. She will only be severely punished if she breaks the laws in force.
“Chinese legislation is halfway between that of the European Union (EU) and that of the United States, with the EU having the strictest approach and the United States the most flexible,” the agency said. AFP Angela Zhang, professor of law at the University of Hong Kong.
A bonus for innovation
The first version of the text aimed in particular to maintain the strict control of the censors on the contents, affirms Jeremy Daum, specialist in China at the faculty of law of Yale (United States). But several restrictions on generative AI have been relaxed. “Many of the strictest controls are abandoned, in favor of promoting development and innovation in the field of artificial intelligence,” he points out on his blog. China Law Translate.
The scope of the rules has been significantly reduced. They now only apply to generative AI available to the general public, and no longer to uses for research and development purposes. “This change can be seen as a sign that Beijing subscribes to the idea of a race for artificial intelligence, in which it must remain competitive,” said Jeremy Daum.