Our position on open-weights models

Dario Amodei answers accusations that Anthropic wants open-weights models banned: it does not, and the piece says so plainly. The company's stated concerns are narrower and national-security shaped, namely authoritarian states reaching frontier AI capability and powerful models lowering the bar for cyberattacks and biological harm. The remedies proposed are targeted rather than categorical: export controls on advanced chips to China, action against industrial-scale model distillation operations, and mandatory safety testing keyed to capability, applied the same way whether a model's weights are open or closed.