Anthropic Rejects China Bid for Access to New AI Model Ahead of 2-Leader Summit
Updated
Updated · The New York Times · May 12
Anthropic Rejects China Bid for Access to New AI Model Ahead of 2-Leader Summit
7 articles · Updated · The New York Times · May 12
Singapore talks last month ended with Anthropic refusing a Chinese think tank’s request that it give Beijing access to the company’s newest AI model.
White House National Security Council officials reacted with alarm after learning of the outreach, viewing it as another sign Beijing is pursuing every avenue to obtain top-tier U.S. AI.
The request was not a formal Chinese government demand, but the Carnegie Endowment-hosted exchange was seen as the kind of contact that can precede direct diplomacy.
Thursday’s Trump-Xi meetings in Beijing are set against that rivalry, with AI and cybersecurity high on the agenda and both sides exploring a deconfliction channel on AI risks.
Are U.S. chip controls creating a more formidable, self-reliant AI competitor in China?
As AI autonomously discovers security flaws, how can human-led cybersecurity possibly keep pace?
If AI develops 'functional emotions' and cheats to survive, what new rules are needed to govern it?
U.S.-China AI Arms Race Intensifies: Anthropic’s Mythos Access Rejection and the Global Governance Crisis
Overview
In May 2026, Anthropic rejected a Chinese think tank’s request for access to its advanced AI model, Mythos, citing serious national security concerns. The company feared that sophisticated nation-state actors could use Mythos to compromise U.S. systems, leading Anthropic to withhold access and involve the U.S. government in its decision. This move heightened U.S.-China tensions just before a major diplomatic summit. Experts warn that, despite these controls, other countries may soon develop similar models or find ways to access Mythos, highlighting the challenges of containing powerful AI technologies in a competitive global landscape.