Dario Amodei, CEO of Anthropic, took to a blog post on Monday to push back against growing criticism that his company wants to crack down on open-weight AI models. He insisted Anthropic has never advocated for an outright ban on such models, which users can freely download, tweak, and run on their own hardware. The statement comes after a coalition of tech giants—including Nvidia, Microsoft, Meta, and Palantir—sent a letter urging policymakers not to impose premature restrictions on open-weight models.

Open-weight models have become a flashpoint in the AI industry, especially as Chinese startups now dominate that space. Some U.S. officials have debated whether to ban or restrict those foreign models. Anthropic, whose Claude models are proprietary and sold to businesses, did not sign onto that recent industry letter—unlike rival OpenAI, which added its name after publication. Amodei used his blog post to clarify his company’s actual position.

He wrote that instead of banning open-weight models as a category, the focus should be on keeping powerful chips away from authoritarian governments, stopping industrial-scale distillation, and requiring safety testing for all sufficiently capable models—open or closed. Distillation, a training method where a new model learns from a larger one, has been a concern: Anthropic recently accused China’s Alibaba of launching the largest known distillation attack against it. Amodei said he agrees with the Friday letter’s call for targeted legal frameworks to address unlawful distillation, and he supports the idea that open-weight models give customers more control and boost competition. But he disagrees that they necessarily favor defenders over attackers or make safety easier. Still, he made it clear: “Protectionist bans would not address my most serious national security concerns.”

Source: CNBC