Anthropic opposes ban on open-weights AI models

Anthropic is strategically pushing back against accusations of regulatory capture by reframing the open-weights debate away from model licensing and toward hardware controls and empirical safety evaluation.

• Banning open-weights usage by US businesses is ineffective at stopping malicious actors or foreign state adversaries who operate outside domestic regulations.

• Prioritizing compute hardware export restrictions and distillation crackdowns directly targets the key bottlenecks that enable state actors to narrow the frontier AI capability gap.

• Universal safety testing for high-capability models offers a pragmatic framework that evaluates empirical risk prior to release rather than assuming risk based on open or closed distribution models.