China slams Anthropic CEO for 'fearmongering' over its AI progress

AI safety debates just collided head-on with superpower geopolitics.

On Monday, China's Ministry of Foreign Affairs publicly hit back at Anthropic CEO Dario Amodei, accusing him of "fearmongering" over Beijing's artificial intelligence ambitions. The diplomatic dust-up follows an extensive essay Amodei published over the weekend, where he warned that a "Chinese lead in AI would pose grave danger for the United States and the world" and urged Washington to double down on export bans for high-end chips and semiconductor equipment.

Amodei went even further, floating a wild scenario where an autonomous swarm of AI agents could overrun the internet within six to 12 months unless frontier labs hit the brakes. His proposed fix? A global pacing agreement that somehow includes China, which he labeled the autocratic state with the most advanced AI capabilities.

Containment wrapped in a safety blanket

Beijing was not amused. Ministry spokesperson Guo Jiakun pushed back during a press briefing, arguing that confrontation and vicious competition only derail global AI governance.

State-run outlet Global Times went straight for the jugular, publishing an editorial that described Amodei's essay as a "Cold War playbook" disguised as a mission for global safety. The timing also was not accidental. Just days earlier, the FBI, NSA, and CISA dropped a joint advisory accusing Chinese developers of "maliciously distilling" capabilities from American flagships like Claude and GPT. China's Commerce Ministry brushed off the report as groundless, calling distillation standard industry practice and accusing the U.S. of chasing an outright tech monopoly.

Meanwhile, Donald Trump took his own swing at Amodei's thesis from the other side, dismissing calls to slow down AI development as a "SICK conspiracy" designed to let China win the race.

Mutual paranoia reaches a boiling point

While Beijing dismissed Washington's warnings, Chinese officials have plenty of AI anxieties of their own.

On Sunday, Chen Yixin, head of China's Ministry of State Security, penned an essay warning that advanced systems from the likes of Anthropic and OpenAI could be weaponized by hostile forces. He argued that foreign actors can now use AI to discover software vulnerabilities at scale and execute complex cyberattacks against China's critical infrastructure while siphoning national data.

Both superpowers are terrified of tech chokepoints, according to Lizzi C. Lee of the Asia Society Policy Institute. As Lee pointed out, if the U.S. demands Beijing cooperate on frontier safety while simultaneously choking its access to frontier compute, that partnership is dead on arrival.

Trump and Chinese leader Xi Jinping are scheduled to discuss AI governance at an upcoming summit on September 24.

Why it matters

This spat confirms that frontier AI has completely outgrown the world of research and is now fully weaponized statecraft.

For builders and enterprises, this ideological trench warfare means a permanently bifurcated tech landscape. Chinese models are already delivering near-frontier performance at a fraction of the cost, largely driven by open-source innovation and heavy model distillation. If your product relies heavily on cross-border infrastructure or frontier APIs, geopolitical compliance is about to become your biggest technical headache.

We think Amodei backed himself into a strategic corner. Trying to argue for tight compute blockades while simultaneously asking the blocked party to honor an international pause button is simply incoherent. Once you strip a geopolitical rival of their computing pipeline, you destroy any leverage you had to bring them to the governance table.

Instead of cooling the temperature, Silicon Valley's apocalyptic warnings are backfiring. In Washington, Trump is weaponizing those exact fears to accelerate development rather than regulate it. Frontier labs might want global coordination, but politicians on both sides are only interested in winning the sprint.