The Cyber Administration of China, responsible for internet supervision in the country, has released the final guidance on generative artificial intelligence (AI), which will come into effect on August 15, 2023.
The rules, titled "Temporary Regulation Methods for the Regulation of Generative Artificial Intelligence Service Systems," are designed to promote the healthy development and regulated use of generative AI, taking into account national security and public interests.
The final version of the rules differs from the initial draft published in April, with some overly strict provisions and restrictions being removed. A softer approach to regulating generative AI has been adopted in the final version:
- The rules will only apply to the use of AI technologies for providing services that generate text, images, audio materials, video, and other content for the population within the territory of China.
- The rules no longer impose penalties for violations related to generative AI. The previous fine, which could reach up to 100,000 yuan (approximately $14,000), has been canceled.
China welcomes innovative applications of AI technologies in various industries and fields. However, the provision and use of generative AI systems must comply with legislative, administrative, social, and moral norms. The rules prohibit the generation of content related to:
- Undermining state power, overthrowing the regime, causing harm to the security, interests, image, and territorial integrity of the country.
- Propagating terrorism, extremism, interethnic hostility, national discrimination, and violence.
- Spreading erotic and pornographic materials, as well as dangerous and inaccurate information.
The rules acknowledge the importance of intellectual property rights in training data and prohibit the use of algorithms, data, platforms, and other advantages for engaging in unfair competition practices, including:
- All training data and basic models must have a legal origin.
- AI must not infringe upon the rights of others to intellectual property objects.
- When using personal data, AI must obtain consent for their processing from individuals.
- AI providers must respond to requests from individuals for checking and correcting information collected and used by AI.
- All content created by AI must be appropriately labeled.
China, along with the EU and the US, is seeking to develop an approach to AI regulation that strikes a balance between technological development, attracting investments, and safeguarding social values and national interests.
At present, American AI startups appear more attractive to investors. However, the new "softened" rules in China may contribute to the development of AI technologies in the country and attract more investments.
For more details, please refer to the "Temporary Regulation Methods for the Regulation of Generative Artificial Intelligence Service Systems" document:
Nadmitov, Ivanov and Partners Law Firm advises on the regulation of artificial intelligence as well as matters related to personal data, information security, and electronic services.
Tel.: +7 (495) 649-87-12