Anthropic's position on open-weights models

Logged at IST: 2026-07-28 18:18 IST

What it is: Dario Amodei lays out Anthropic's stated position on open-weights models amid discussion of possible restrictions on Chinese open-weights models.

Gist: The post explicitly says Anthropic has never advocated for a ban on open-weights models as a category. Dario says open-weights models without dangerous capabilities are a public good, and that protectionist bans on US business use would not address his main national-security concerns.

Those concerns are split into two categories: authoritarian governments building models more powerful than US models and using them for military superiority or repression, and powerful models being misused for cyberattacks, biological attacks, or serious alignment failures. On the second point, he argues open weights can be riskier because guardrails and monitoring are hard once weights are released, but a US business-use ban would not stop bad actors.

Anthropic's preferred interventions are narrower: block powerful chips and chipmaking equipment from reaching China, crack down on industrial-scale distillation, and require safety testing for all sufficiently capable models, open and closed. Dario agrees open weights can expand access, competition, and customer control, but rejects assuming that broad access necessarily helps defenders more than attackers.

Newsletter angle: Useful policy clarification because it separates “ban open weights” from Anthropic’s actual ask: constrain frontier compute, curb distillation, and test dangerous capability thresholds directly.