China published the proposed security requirements for companies providing services based on generative artificial intelligence. The document indicates a black list of sources that cannot be used to teach AI models.
Generative AI, which has become popular thanks to the success of the Chatbot from Openai, is learning on the basis of past data and creates new content, such as text or images.
The requirements were published on Wednesday by the National Committee for Standardization of Information Security. The Committee includes representatives of the Cyberspace Administration of China, the Ministry of Industry and Information Technologies, as well as the police.
The Committee proposes to assess the security of each content used to teach publicly available generative AI models. Content containing “more than 5% of illegal and harmful information” will be blacklisted. Such information includes calls for terrorism, violence, “overthrow of the socialist system”, “undermining the image of the country” and “national unity.”
The draft Rules also states that the information that has been censored in the Chinese Internet should not be used to teach models.
The publication of these requirements took place a month after the regulators allowed several Chinese technological companies, including the Baidu search giant, to launch their chatbots based on generative AI.
Since April, the Cyberspace Administration has announced the need to provide security assessments before launching services based on generative AI.
In July, the cyberspace regulator published measures governing such services that, according to analysts, were much less strict than the measures presented in the April project.
The requirements published on Wednesday are obligated by organizations that teach these models of AI to receive consent of persons whose personal information, including biometric data, is used for training purposes.
The document also contains detailed recommendations for the prevention of intellectual property violations.
Many countries of the world are faced with the task of establishing frames for this technology. China sees AI as an area in which the country wants to compete with the United States and sets itself the goal of becoming a world leader in this area by 2030.