Anthropic CEO Dario Amodei directly refuted claims that his company supports banning open-weights AI models, outlining a more targeted policy approach in a public statement on July 27.
The executive pushed back against accusations that Anthropic wants sweeping restrictions on open-weight releases, describing such models as helpful when they lack dangerous capabilities. The clarification dropped into a heated policy debate as US officials weigh limits on American companies using Chinese open-weight models, prompting a counter-letter from industry leaders defending the practice.
Amodei flagged two nightmare scenarios: authoritarian states building frontier AI that surpasses American capabilities for military or surveillance use, and the potential for powerful models to enable cyberattacks or biological weapons. He contended that restricting open-weight usage by US businesses misses both threats entirely.
“Bad actors are unlikely to be legitimate US businesses,” Amodei wrote. Instead, he proposed three focused measures: blocking the sale of advanced chips to China and cracking down on smuggling, targeting industrial-scale distillation operations that let China evade chip restrictions, and mandating safety testing for all sufficiently capable models regardless of origin or openness.
The post was widely seen as an attempt to reset the narrative after some in the tech industry accused Anthropic of using safety concerns as cover for anti-competitive practices.
