Our position on open-weights models
Points and comments are a snapshot, not live.
Anthropic's CEO Dario Amodei clarifies the company does not advocate banning open-weights AI models.
Amodei states Anthropic has never advocated for a ban on open-weights models, calling those without dangerous capabilities a public good. He outlines two nightmare scenarios: authoritarian states building superior AI for military or repression, and misuse for cyber or biological attacks. To address these, he supports not selling powerful chips to China, cracking down on industrial-scale distillation, and requiring mandatory safety testing for sufficiently capable models (both open and closed). He notes that banning Chinese open-weights models from US businesses would not stop bad actors and would only protect US AI companies from competition.
What commenters are saying
Many commenters are skeptical of Anthropic's motives, accusing the company of using national security rhetoric to justify anti-competitive regulation. Several call the position hypocritical, noting Anthropic itself trains on vast amounts of data while opposing distillation. A smaller camp finds the position reasonable and consistent. Specific complaints include that required safety testing amounts to a de facto ban, and that the focus on China ignores the US government's own authoritarian potential. One commenter argues the real danger is not hacking or bioweapons but autonomous drones used for mass surveillance and infrastructure attacks.