China advances AI full-lifecycle legislation: Companies must proactively disclose risks, and a regulatory sandbox will be established for trial and error.

Beating AI News Flash, November 3 - The National Development and Reform Commission stated that it will accelerate the advancement of comprehensive AI legislation, establishing safety standards and ethical guidelines covering the entire lifecycle from data, computing power, and algorithms to specific applications, including R&D and deployment. AI developers and operators will also be urged to promptly disclose safety risks and proactively report incidents and vulnerabilities.

While regulation tightens, enterprises will still be given room for trial and error. The new rules will set a reasonable transition period and support the establishment of "regulatory sandboxes" across different industries, allowing companies to test new technologies in controlled environments. The sandboxes will be supported by application, testing, use, and exit mechanisms, and regulatory rules will be periodically evaluated and adjusted in a timely manner.

Beyond model and technical safety, regulators will also focus on AI's impact on people, including monitoring the risk of AI displacing jobs, improving social security, and preventing users from becoming addicted to AI or developing emotional dependence. Deepfakes, algorithmic discrimination, and privacy breaches are also within the scope of governance.

China introduced measures for managing public-facing generative AI services as early as 2023, and in May this year added comprehensive legislation to the State Council's work plan. Regulatory sandboxes have also been previously explored in some AI service sectors.