China to accelerate comprehensive AI law and safety standards

From guidance to formal law

China’s National Development and Reform Commission (NDRC) announced plans on November 3 to accelerate the creation of a basic, comprehensive AI law. This move shifts the regulatory landscape from temporary technical guidance toward a permanent legal framework. The new rules will aim to establish safety standards and ethical guidelines that cover the entire life cycle of artificial intelligence.

According to the NDRC’s official transcript, the legislation will specifically target data, computing power, and algorithms. It will also address key application scenarios to ensure that AI development aligns with state goals. The agency noted that these rules will be reviewed periodically so that provisions can be adjusted if they no longer fit the evolving technology.

This legislative push follows a rapid period of technical updates from the national cybersecurity standards committee, known as TC260. By our arithmetic, the committee released three different versions of its AI Safety Governance Framework in just 24 months. The NDRC's new plan seeks to codify these types of standards into a unified legal structure for the country.

New duties for AI developers

A significant new element of the NDRC's plan is the specific duty placed on AI developers and operators. The government will push these companies to strengthen their risk management practices. Specifically, firms will be expected to disclose safety risks promptly and proactively report any safety incidents or vulnerabilities to the authorities.

As reported by Bloomberg, these rules are expected to reach major developers such as DeepSeek, Moonshot, and Alibaba. The goal is to create a system where the state is informed of potential hazards before they escalate. This transparency is intended to help regulators manage the risks associated with large-scale model deployment.

To balance these requirements, the NDRC stated it will provide reasonable transition periods for companies to come into compliance. The government also plans to offer sector-specific guidance and build evaluation and certification services. These measures are designed to help reduce the overall compliance costs for firms as they navigate the new legal requirements.

Sandboxes and economic impact

The NDRC plans to support AI regulatory sandboxes for various industries. These sandboxes provide a controlled environment where companies can test, use, and apply new AI technologies under specific rules before a full market release. The government will establish clear guidelines for how companies can enter, operate within, and eventually exit these testing zones.

Beyond technical safety, the government is looking at the broader economic and social implications of AI. The NDRC intends to monitor the risk of job displacement caused by automation. They plan to strengthen early warning systems for these changes and link social insurance to the evolving ways in which people work.

Furthermore, the government will update the national occupational classification to include roles like model trainers and data annotators. This reflects an effort to recognize the new types of labor being created by the AI industry. By formalizing these roles, the state aims to better integrate AI work into the national labor and social security systems.

Global safety and open source

The push for safety standards is linked to global concerns regarding AI security and cyber-breaches. Wang Ruomeng, a deputy director-general of the NDRC, noted that several AI security incidents have occurred worldwide since the start of the year. In response, China is seeking international cooperation on vulnerability notification and risk warning systems.

The NDRC highlighted a recent agreement between China and the United States to establish a dialogue and a communication channel for AI incidents. This follows a September meeting where both sides agreed to coordinate on safety governance. Wang emphasized that AI safety is a global challenge that no single country can solve alone.

Despite the move toward stricter safety laws, China intends to maintain its commitment to open-source development. Wang stated that the country will adhere to open source and inclusive development. This approach is seen as a key part of the playbook for the global spread of Chinese models, which the government continues to encourage through 2030.