Anthropic CEO Denies Advocating for Ban on Open-Weight AI Models

Anthropic CEO Dario Amodei stated on Monday that his company has "never advocated for a ban on open-weights models," addressing growing criticism within the technology industry. Amodei published his clarification in a blog post, seeking to counter perceptions that Anthropic is pushing for excessive control over AI development.

The statement follows a letter released last week by a coalition of tech companies, including Nvidia, Microsoft, and Meta. The letter urged policymakers to avoid "premature restrictions" on open-weight models, which can be downloaded and modified by users. While rival OpenAI signed the letter, Anthropic notably did not.

Anthropic primarily develops proprietary models like its Claude family, which are sold to businesses. Amodei reiterated that Anthropic's position is not to ban open-weight models as a category. Instead, he proposed focusing on preventing authoritarian actors from accessing powerful chips, curbing industrial-scale "distillation" (where smaller models are trained using outputs from larger ones), and requiring safety testing for all capable models, regardless of their openness.

While acknowledging that open-weight models offer users greater control and foster competition, Amodei disagreed with the notion that they inherently favor cyber defenders over attackers or simplify the development of safeguards. He affirmed that protectionist bans would not address his most serious national security concerns.