China's government and its leading AI labs have rejected Anthropic CEO Dario Amodei's proposal that Washington and Beijing cooperate to slow frontier AI development, calling the pitch a thinly disguised attempt to freeze the current US lead in place. On Monday, foreign ministry spokesperson Guo Jiakun labeled the plan 'fear-mongering, confrontation, and vicious competition.' The rebuke came two days after Amodei published a blog post titled 'We Must Pace the Frontier' arguing that autonomous AI risks warrant urgent bilateral coordination.
The disagreement is not about whether advanced AI carries risk. Chinese officials, researchers, and industry figures acknowledge the danger. Their objection is structural: Amodei made the slowdown conditional on the US widening its capability gap over China, including tighter chip export enforcement and measures to prevent Chinese firms from distilling US models.
“Global pacing will require cooperation with China, the autocratic country with by far the most advanced AI capabilities.”— Dario Amodei, Anthropic CEO
Amodei called for the US and other democracies to maintain 'as large as possible' a lead over autocracies, then argued that weakening China's AI industry would somehow make a future agreement more likely. He did not explain the mechanism. In a subsequent television interview, Amodei conceded he has 'no idea how to negotiate with China.'
Key facts
- 01Anthropic CEO Dario Amodei published 'We Must Pace the Frontier' on Saturday, calling for US-China cooperation to slow AI development.
- 02China's Ministry of Foreign Affairs spokesperson Guo Jiakun called the proposal 'fear-mongering, confrontation, and vicious competition' on Monday.
- 03Amodei tied any slowdown to keeping the US lead over China 'as large as possible' through chip export enforcement and anti-distillation measures.
- 04A DeepSeek researcher published a Sunday essay rejecting the idea of Anthropic or OpenAI controlling frontier AI.
- 05Preliminary US-China AI safety talks are underway ahead of Xi Jinping's planned Washington visit later this month.
Kevin Xu, founder of Interconnected Capital and a former White House staffer under President Barack Obama, said the framing is the core problem. Alvin Wang Graylin, a senior fellow at the Asia Society Policy Institute's Center for China Analysis and a digital fellow at the Stanford Institute for Human-Centered AI, put it more bluntly: the pacing proposal reads as inauthentic to Chinese audiences because its stated purpose is to widen the gap.
The Chinese ministry followed Guo's press conference with a statement that 'China puts equal emphasis on development and security in terms of AI.' The framing matters. Beijing wants to be treated as a governance partner and rejects being cast simultaneously as an essential counterparty and an adversary requiring containment.
Beijing's threat model also differs from the one that dominates Silicon Valley safety discourse. Michelle Nie, a visiting fellow at the Center for a New American Security, said Chinese policymakers are skeptical of AGI timelines and doubtful of extinction-risk narratives. Chinese leaders are more focused on AI-driven disinformation, cyberattacks, espionage, and social unrest — threats to Chinese Communist Party rule rather than to humanity.
Chen Yixin, China's minister of state security, laid out that framework in a recent essay warning of political disinformation, cyberattacks, espionage, social unrest, and new military threats from AI. Doomsday scenarios that dominate US frontier lab discourse got far less attention. Nie said Chinese policymakers tend to read US safety appeals as instrumental — meant to inflate valuations ahead of IPOs or slow down Chinese competitors.
That skepticism runs deep in China's AI industry, which has focused on lower-cost open-weight models. In a widely shared Sunday essay, a researcher at DeepSeek argued that AI will either lift living standards broadly or entrench a handful of tech companies that control the frontier and the resources around it.
Not all channels are closed. Preliminary US-China negotiations are already underway ahead of Xi Jinping's planned Washington visit later this month to meet President Donald Trump. Both governments have flagged overlapping concerns: cybersecurity threats, misuse of powerful models, and unexpected behavior from increasingly capable systems. Nie said narrow, verifiable agreements remain possible if both sides can design credible compliance mechanisms.
Kristy Loke, a research fellow at the AI safety nonprofit MATS Research, argued the Anthropic pitch may set back the very cooperation it invokes. She said Amodei is undermining one of the few available levers for reducing the risks he claims to take seriously.
The pattern here matters for anyone tracking how AI diplomacy actually gets done. Frontier labs increasingly want two things from Washington at once — export controls that constrain competitors, and a diplomatic posture that treats those same competitors as partners on safety. Beijing has now made explicit what the contradiction costs. Any US-China AI safety framework that survives the next twelve months will be built by governments negotiating narrow, verifiable commitments, not by lab CEOs asking one side to hobble itself first and negotiate later.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
