Anthropic CEO Dario Amodei Backs AI Safety Tests Over Blanket Ban on Open-Weight Models

Anthropic CEO Dario Amodei has said the artificial intelligence company has never supported banning open-weight AI models, pushing back against criticism that it wants tighter restrictions on the technology.

In a blog post published on Monday, Amodei said Anthropic's priority is to ensure powerful AI systems are developed safely through mandatory testing, while also preventing technologies that could help China strengthen its AI capabilities.

His remarks come days after Anthropic declined to join an industry letter urging US policymakers not to impose broad restrictions on open-weight AI models, triggering speculation that the company favoured a ban.

Never Advocated for a Ban

Amodei said suggestions that Anthropic wants open-weight AI models to be prohibited are incorrect. He said the company has "never advocated for a ban" on such models and continues to recognise the benefits they offer, including wider access to AI and greater competition.

However, he argued that open-weight models become harder to monitor once they are publicly released, making it more difficult to prevent misuse by malicious actors.

Instead of restricting open-weight AI, Amodei said policymakers should focus on three areas: limiting access to advanced AI chips by authoritarian governments, stopping large-scale AI distillation and introducing mandatory safety testing for highly capable AI models, regardless of whether they are open or proprietary.

Different View From Industry Peers

Anthropic's position differs from that of several major technology companies, including Nvidia, Microsoft, Meta, OpenAI and Palantir, which recently signed a joint letter asking Washington to avoid what they described as premature restrictions on open-weight AI systems.

The signatories said open-weight models improve access to advanced AI, increase customer choice and encourage innovation across the technology ecosystem.

Although Anthropic stayed out of the letter, Amodei said he agrees with some of its broader goals. However, he disagreed with the argument that open-weight AI automatically makes systems safer or gives defenders an advantage against cyber threats.

China Remains a Key Concern

Amodei also renewed his call for stronger measures to prevent China from gaining access to advanced AI capabilities.

One of the key issues, he said, is AI distillation, a technique in which developers use outputs from a more advanced AI model to train another system. Anthropic and OpenAI have previously alleged that some Chinese AI companies used the method to replicate the capabilities of their models, allegations that China has denied.

The issue has gained greater attention following the emergence of Chinese open-weight AI models such as Moonshot AI's Kimi K3, which has intensified discussions in Washington over how the US should respond to China's rapid progress in artificial intelligence.

AI Rules Under Discussion

The debate comes as the Trump administration prepares a framework for regulating advanced AI systems.

US officials are weighing proposals that include mandatory safety testing for frontier AI models, while also examining measures to curb the alleged misuse of American AI technology by overseas companies.

As policymakers work on the new framework, Amodei said the focus should remain on managing the risks posed by advanced AI systems rather than imposing a blanket ban on open-weight models.