Anthropic's Amodei: Open-Weight AI Not the Problem, China's AI Is

Dario Amodei, co-founder and CEO of Anthropic, addressed industry speculation on Monday afternoon regarding his company's alleged support for U.S. government initiatives to ban open-weight AI models, specifically those from China, or open-weight models in general.

“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,” he explicitly stated, emphasizing his position.

Amodei's clarification followed an open letter shared by Nvidia founder and CEO Jensen Huang on X (marking his inaugural post on the platform) on Friday. This letter, co-signed by Nvidia and a diverse group of prominent AI companies including Hugging Face, Meta, Microsoft, and Mistral, urged policymakers to refrain from imposing sweeping “premature restrictions” on open-weight AI models.

Although the letter did not explicitly name China, the broader industry discourse has frequently revolved around accusations that Chinese AI laboratories are advancing their capabilities, often through the alleged theft of intellectual property from their American counterparts. A common method cited is "distillation," a process where one AI system repeatedly queries another to reverse-engineer its operational mechanisms.

In a blog post published Monday, Amodei articulated that he categorizes open-weight models distinctly from the potential threat posed by China. He asserted, “Open-weights models that don’t have dangerous capabilities are a public good: they don’t cost anything besides the compute needed to run them, and they provide value to businesses, developers, and researchers.”

Amodei further clarified in his post that while he harbors long-standing concerns about AI, the use of open-weight models by businesses, even those originating from China, does not fall within the scope of these fears.

Instead, his primary apprehension centers on the prospect of “authoritarian governments” developing AI models more powerful than those in the U.S. with the aim of achieving “permanent military superiority” or employing AI for the repression of their own populations. He noted that while the Chinese Communist Party (CCP) is not his sole concern among authoritarian regimes, it is undeniably the “most capable.”

Amodei also voiced concerns that AI could facilitate “biological attacks,” extending beyond mere cybersecurity threats. In such scenarios, he believes open-weight models pose greater dangers due to the inherent difficulty in implementing effective guardrails or monitoring their usage. Citing a report from the UK AI Security Institute, he highlighted that once open-weight models are released, their recall becomes impossible.

This perspective contrasts sharply with the arguments put forth by open-source advocates, who contend that broad access to powerful open models, not controlled by any single entity, actually empowers defenders to better protect themselves.

To counter potential threats from China, Amodei proposed several actions, including restricting its access to powerful chips—a measure that aligns with existing U.S. policy—and advocating for a formal crackdown on distillation. The U.S. has previously threatened sanctions against China should it confirm instances of intellectual property theft involving American AI models.

Intriguingly, Amodei also expressed support for growing international efforts, some led by the U.S., to establish an AI model safety testing organization, particularly if global participation, including China's, could be secured.

“I think this idea is actually close to a consensus: I have been heartened both that the Trump administration has moved in this direction in recent months,” Amodei wrote. He added, “Note that to be effective, testing would need to be global, which means even the CCP would need to be on board. I think this may actually be possible … limited cooperation around preventing AI biological weapons may be possible because it is in China’s interest too.”