On September 12–13, 2026, Anthropic CEO Dario Amodei proposed slowing AI development and allowing independent evaluators to access AI labs. China’s Foreign Ministry rejected this proposal on September 14, 2026, labeling it as fearmongering and a strategy to hinder China’s technological progress.
Beijing views independent safety monitors as a potential security risk, fearing they could serve as a backdoor for espionage. Additionally, the Chinese state security minister stated on September 13–14 that Anthropic’s AI models pose threats to critical infrastructure in China.
Instead of external oversight, China is developing a national standard for AI agent safety. This framework mandates state-coordinated testing, where developers must meet government-defined benchmarks under direct state supervision.
The disagreement precedes bilateral AI governance talks scheduled for late September, including a meeting between Donald Trump and Xi Jinping on September 24. This highlights the growing gap between Western calls for international oversight and China’s focus on state control.
AI companies now face two distinct regulatory regimes. Firms operating in both jurisdictions must navigate conflicting requirements, as satisfying state-coordinated standards in China may contradict Western demands for independent evaluation.
