Anthropic has publicly clarified its position on open-weights AI models, stating it has never called for their prohibition and considers models without dangerous capabilities to be a public good. The company instead advocates for targeted measures including export controls on powerful chips, action against industrial-scale distillation, and safety testing for all sufficiently capable models.
Open-weights models that don’t have dangerous capabilities are a public good.
The clarification follows controversy over a letter circulated by Nvidia CEO Jensen Huang supporting open-weight models. Anthropic declined to sign, drawing criticism. Google and OpenAI signed the letter after initially abstaining.
Anthropic CEO Dario Amodei acknowledged that open-weight models expand access and competition but warned they could enable biological attacks. He supported the creation of a global AI safety testing organization that includes China.
Protectionist bans would not address my most serious national security concerns.
White House AI adviser David Sacks accused Anthropic of using safety concerns to protect its business model. Anthropic has denied this, with Amodei stating that goal has never been his aim.