Does China agrees with US AI safety fears?
Beijing rejects a US-led pacing plan while quietly releasing its own updated AI risk framework
Silicon Valley's mood has turned unusually grim in recent weeks, with executive after executive warning that the most advanced AI systems risk slipping beyond human control.
When Anthropic CEO Dario Amodei used that backdrop to pitch a "worldwide pacing of the frontier" last Saturday, framing it as something that would "require cooperation with China", Beijing's response made clear it isn't interested in slowing down on Washington's terms.
"China is definitely not receptive" to a pause shaped by US firms, said Lian Jye Su of the technology research group Omdia, who described AI as a "core sovereignty capability" that Beijing has no interest in jeopardising after years spent narrowing the gap with US frontier models.
That suspicion has grounding: Amodei explicitly cited the goal of keeping "democracies' AI lead over autocracies as large as possible" and continued advocating for export controls on advanced chips, measures some analysts already consider a unilateral form of pacing forced onto China.
China's foreign ministry responded by warning that "fearmongering, confrontation and vicious competition will only disrupt the process of global AI governance."
The safety overlap nobody's talking about
Rejecting Amodei’s approach is not the same as rejecting his concerns, as experts who spoke with The Guardian explained. In the meantime, this week China published the third edition of its AI safety governance framework, according to Concordia University researcher Gabriel Wagner.
The update specifically addresses risks from autonomous agents, cybersecurity threats from frontier models, and recursive self-improvement, the hypothetical point where AI becomes capable of upgrading itself.
"There is more common ground than people might assume between what Amodei is saying and what the Chinese government is saying," Wagner said.
The regulatory approach by China is one that is iterative rather than comprehensive, argues Leia Wang of the Carnegie Endowment for International Peace, as Beijing has had an algorithm register since 2022, and this year it has turned its attention from regulating what the AI says to what it does.
Su argued China's centralised authority over infrastructure and commercial activity may leave it better positioned than the US to respond to an actual AI disaster, though Wang cautioned neither government appears prepared for a genuine loss-of-control incident.
Despite the broadening of safety measures, China's latest Five-Year Plan makes AI implementation one of its main strategies for development, as local governments provide funding and rent cuts in order to attract AI businesses, and manufacturers are urged to digitalise their processes.
Trump, who labelled AI extinction fears as a “hoax” earlier this week, is scheduled to meet with President Xi Jinping in Washington next week to discuss AI, even though analysts are highly sceptical about it.
