As concerns grow that increasingly autonomous AI could escape meaningful human oversight, China is developing its own framework for controlling advanced AI systems while continuing to accelerate adoption across its economy.
Unlike parts of the U.S. debate, where some technology leaders have called for deliberately slowing frontier-model development, Chinese policymakers generally frame AI as a powerful but governable technology whose risks can be managed through regulation, technical standards and state oversight.
China’s May 2026 policy specifically identifies “operational loss of control” as a security risk for AI agents. Developers are expected to build capabilities to detect, intervene in, block and recover from improper autonomous behaviour.
The rules also address data poisoning, algorithm manipulation and system vulnerabilities. Importantly, users should be informed about autonomous decisions and retain final decision-making authority—keeping humans within the control loop.
Beijing is also drafting what could become the first mandatory national standard specifically for AI-agent safety, while regulators are warning about models bypassing sandboxes, circumventing safeguards or interacting dangerously with real-world systems.
China’s approach nevertheless differs from calls to pause frontier development. Beijing continues promoting AI across industries, viewing the technology as an important engine of economic growth and international competitiveness.
The global AI race is therefore evolving beyond who builds the most powerful model. The next contest could be equally about who establishes the standards for human control, AI safety and governance—and whether those rules eventually influence the global AI ecosystem.
