Anthropic CEO Dario Amodei doesn’t advocate banning openweight models

Anthropic CEO Dario Amodei said on Monday that the company has “never advocated for a ban on open-class models,” seeking to repel criticism that has surfaced across the tech industry that the AI ​​Institute is trying to exercise too much control over the future of artificial intelligence.

Amodei announced his views in a blog post on Monday after a consortium of tech companies including Nvidia, Microsoft, Meta and Palantir released a letter late last week urging policymakers to avoid “premature restrictions” on the open-weight model, which users can download, modify and run on their own infrastructure. Chinese startups now dominate the market, and some government officials are starting to consider whether these models should be banned or restricted in the United States.

Anthropic is best known for developing its own family of models called Claude, which it sells to businesses. Its main rival, OpenAI, also primarily builds closed models, but the company signed an endorsement after the publication of the open weight letter, while Anthropic did not.

“To summarize my and Anthropic’s position, we do not and do not advocate a ban on indiscriminate weight models as a category,” Amodei wrote. “Instead, we should focus on keeping powerful chips out of the hands of authoritarians, stopping industrial-scale distillation, and mandating safety testing of all fully functional models, open and closed.”

Distillation is an AI training method that uses the output from larger models already on the market to build smaller, less powerful models. Last month, Anthropic sent a letter to the U.S. Senate Banking, Housing and Urban Affairs Committee alleging that China’s Alibaba, which developed the Qwen family of models, had launched “the largest distillation attack ever known” against the company.

In Friday’s letter, the signatories said concerns about illegal distillation should be addressed through a “targeted legal and commercial framework,” a view echoed by Amodei on Monday.

Amodei said he agreed with other parts of the letter. In short, the open weight model gives customers more control, expands access to the AI ​​economy, and increases competition in some use cases. But he said he disagrees with the idea that the open-weight model favors cyber defenders over attackers or makes it easier for users to develop safeguards.

Still, Amodei made it clear that he does not support banning open weight models.

“A protectionist ban does not address my most serious national security concerns,” Amodei said.

WATCH: Anthropic’s new AI model rivals Fable 5 and is cheaper as companies worry about costs