China Implements Regulations on Generative AI

Yomiuri Shimbun file photo
The Cyberspace Administration of China, which is in charge of regulating generative artificial intelligence, is seen in Beijing.

BEIJING/WASHINGTON — China on Tuesday implemented a set of regulations on the management of generative artificial intelligence, becoming the first major country to introduce full-fledged rules on generative AI.

While Washington is considering regulations that would emphasize the protection of AI users, Beijing primarily aims at controlling AI to prevent it from posing a threat to national security.

Interim measures

The regulations consist of 24 articles, and call for adherence to socialism’s central values, among other things. The articles reflect the government’s strong sense of caution about generative AI’s transmission of views that are different from official announcements, showing a focus on the control of speech.

Under the regulations, generative AI made in the United States and other foreign countries will effectively be eliminated from China.

Apple Inc. pulled more than 100 generative AI-related apps from its Chinese App Store earlier this month. The U.S. company is believed to have done so because the apps, including those powered by ChatGPT — developed by another U.S. firm, OpenAI Inc. — are deemed to be in violation of Chinese regulations.

The regulations include ones for the protection of intellectual property rights and personal information. They prohibit discrimination and the distribution of false information.

Since these elements are also emphasized in draft regulations being considered in the United States and European countries, China apparently aims to keep in step with them. Having positioned the latest regulations as interim measures, China is expected to review them based on how effective they are in practice.

Striking a balance

The regulations also include the encouragement of independent innovation in basic technologies and participation in the formulation of international rules, among other items, in order to strike a balance between restrictions and the promotion of development.

Li Zhihui of Nomura Research Institute, Ltd. said that the regulations, when compared to a draft released in April, were toned down in terms of restrictions and focused more on the importance of development.

The draft had stated that violations of the rules would result in suspension of service, criminal penalties or fines from 10,000 yuan to 100,000 yuan (about ¥202,000 to ¥2.02 million), but those have been removed.

It has been pointed out that China lags behind the United States in terms of generative AI-related technology. China is believed to have been concerned that it might fall further behind in the AI development race if it imposed too many regulations.

U.S. moves

The United States has also been focusing on regulating AI. On July 21, U.S. President Joe Biden met with executives from seven leading U.S. IT companies and agreed to introduce rules to ensure the safety and transparency of AI products. Biden plans to issue an executive order to ensure the agreements’ effectiveness.

The agreements, meant to guarantee the safety and reliability of AI products, include testing AI systems before their release. The testing will be carried out in part by independent experts. The companies are also obliged to share information with the public on managing AI risks.

The companies also committed to developing mechanisms such as a watermarking system to ensure that users know whether audio and video content is AI-generated. These are voluntary commitments by the companies.

To support responsible development of AI technology, the Biden administration released the “Blueprint for an AI Bill of Rights,” which lays out principles to be considered in AI development, in October last year. It calls on companies to take measures to ensure appropriate data collection and protect individuals and communities from discrimination.

The administration also asked Congress to introduce bipartisan legislation that includes restrictions on the collection of personal information.

The Organization for Economic Cooperation and Development is working on new guidelines for generative AI. The OECD is expected to provide principles that member countries, including Japan and the United States, should focus on when creating domestic rules.