Anthropic chief executive Dario Amodei has issued a statement denying the company has ever advocated banning open-weights AI models.
Amodei said open-weights models without dangerous capabilities are a public good, but reiterated two long-standing national security concerns: that authoritarian governments could use superior AI for military dominance or repression, and that open models are harder to monitor or withdraw once released, raising cyberattack and bioweapon risks.
Rather than banning open models, Amodei said Anthropic supports three measures: restricting sales of advanced chips and chipmaking equipment to China, cracking down on large-scale "distillation" operations that let China narrow the AI gap despite chip restrictions, and mandatory safety testing for all sufficiently capable models regardless of origin or openness.
He said he agreed with parts of an industry letter backing open-weights models, including that open weights broaden access and strengthen competition, but disputed its claim that openness necessarily aids defenders more than attackers, citing biological risk as an area where he believes the reverse may be true.
