Anthropic CEO denies seeking ban on open-weight AI models

Anthropic clarifies position on open AI models

Anthropic CEO Dario Amodei published a statement Monday asserting that his company has never called for prohibiting open-weight AI models, responding to mounting industry criticism that the lab seeks to restrict how artificial intelligence develops.

The clarification came after a coalition including Nvidia, Microsoft, Meta, and Palantir issued a letter urging lawmakers to reject "premature restrictions" on open-weight models—AI systems users can download, modify, and operate on their own hardware. OpenAI signed the letter after its publication, but Anthropic notably did not, fueling speculation about the company's position.

"To summarize my and Anthropic's position, we have not and are not advocating for a ban on open-weights models as a category," Amodei wrote in his blog post. Instead, he outlined three alternative priorities: preventing powerful chips from reaching authoritarian governments, halting industrial-scale distillation, and mandating safety testing for all sufficiently capable models regardless of their distribution approach.

Why it matters

The debate over open-weight models carries significant implications for AI competitiveness and national security. Chinese startups currently lead the open-weight market, prompting some U.S. officials to consider restrictions. How policymakers resolve this tension between openness and security will shape which companies can participate in AI development and whether the technology remains accessible to smaller players and researchers.

The distillation dispute

Amodei's statement emphasized concerns about distillation—a technique where developers train smaller models using outputs from larger, existing systems. Last month, Anthropic sent a letter to the Senate Committee on Banking, Housing, and Urban Affairs claiming that China's Alibaba conducted "the largest known distillation attack" against its Claude models to build the Qwen family of AI systems.

The Friday coalition letter acknowledged distillation concerns but argued they should be addressed through "targeted legal and commercial frameworks" rather than broad restrictions on open-weight models. Amodei expressed agreement with this approach.

Points of agreement and disagreement

Amodei said he supports several arguments made in the coalition letter, including that open-weight models provide customers greater control, expand economic access to AI technology, and strengthen competition in certain applications.

However, he disputed claims that open-weight models inherently favor cybersecurity defenders over attackers or simplify the development of safety measures. He also rejected the notion that protectionist bans would address his primary national security concerns.

Anthropic builds the proprietary Claude model family, which it licenses to businesses—a closed approach similar to chief rival OpenAI. The company's business model differs from Meta and other firms that release open-weight versions of their AI systems.

These details were first reported by CNBC.