Anthropic's Amodei backs open-weight AI in principle, but pushes chip controls and anti-distillation measures

What happened

Anthropic CEO Dario Amodei broke his company's silence on open-weight AI policy on July 27, 2026, publishing a statement that sought to dispel growing speculation that Anthropic wanted open-weight models banned.

"Anyone who has read my past writing should know that I don't regard such bans as a useful measure, but let me state it clearly so that there is no doubt: Anthropic has never advocated for a ban on open-weights models," Amodei wrote, with emphasis his own. He added that open-weight models "that don't have dangerous capabilities are a public good."

The statement came days after Nvidia founder and CEO Jensen Huang made his first-ever post on X to share an open letter co-signed by Nvidia, Hugging Face, Meta, Microsoft, Mistral, and others urging policymakers not to impose "premature restrictions" on open-weight AI. While that letter did not mention China specifically, the surrounding industry debate has centered on allegations that Chinese AI labs are gaining capability partly through intellectual property theft — including distillation, a technique where one model extracts knowledge from another at far lower cost than training from scratch.

Amodei outlined three measures he supports instead of a ban. First, enforcing chip export controls and cracking down on the "rampant smuggling" used to get advanced chips into China, arguing that without US chips, scaling laws prevent China from training models more powerful than American ones. Second, stopping industrial-scale distillation operations that allow Chinese labs to build capable models cheaply. Third, mandatory safety testing for all sufficiently capable models — open and closed, regardless of country of origin — with less capable models from startups and academia exempted entirely.

He also pushed back on the idea of a US-specific ban on Chinese models, writing that it "wouldn't fix anything, because bad actors are unlikely to be legitimate US businesses." Such a ban "would protect US AI companies from competition, but that has never been my goal."

Amodei described two "nightmare scenarios" driving his thinking: authoritarian states, especially China, building more powerful models than the US and using them for military superiority or surveillance; and powerful open models being misused for cyber or biological attacks. He argued open weights are riskier than closed models because they lack safeguards and cannot be recalled once released.

Why it matters

The open-weight debate has split the AI industry along lines that don't map neatly onto company categories. Anthropic, which sells closed models, is now on record supporting open weights in principle — but with conditions that could still translate into significant regulation for the most capable models. Meanwhile, Anthropic's call for stricter chip export controls creates an awkward dynamic with its strategic partner Nvidia, which has a commercial interest in selling chips globally, including to China. As Gizmodo noted, Anthropic's stance is "unlikely to fly" with Nvidia on this specific point.

For developers and enterprises building on open-weight models, the policy contours Amodei outlined — mandatory safety testing, anti-distillation enforcement, and chip controls — signal where future US regulation may land. None of these would ban open weights outright, but they could raise the cost and complexity of releasing or deploying the most capable open models.

What to watch

• Whether US policymakers adopt any of Amodei's three proposals, particularly mandatory safety testing for capable models regardless of origin.
• How Nvidia responds to Anthropic's export-control push, given the companies' strategic partnership and Nvidia's interest in global chip sales.
• Whether the open-letter coalition narrows or widens its ask as the policy debate moves from industry statements to legislative proposals.
• How Chinese AI labs' distillation practices evolve if US enforcement efforts intensify.