Anthropic CEO: No Open-Weights Ban, Control AI Chips

Anthropic CEO Dario Amodei rejects a blanket ban on open-weights AI models. He says his company has never advocated for such a ban. Here's the thing: Amodei is pushing for three different policy measures instead. He wants tighter export controls on advanced AI chips. He also calls for legal and commercial deterrents against industrial-scale "distillation" operations. Finally, he proposes mandatory pre-release safety testing for all sufficiently capable AI models. This would apply regardless of whether they are open or closed, or where they were built. What's interesting is that an "open-weights" model is an AI system whose underlying code is publicly available. This is different from a "closed" model, like Anthropic's Claude, which remains under company control. Amodei's statement directly responds to reports about US officials considering restrictions on Chinese open-weights models. He argues that a usage ban targeting US companies would miss the point, as foreign militaries and intelligence services would not be constrained by American business practices. The bottom line is these proposed measures would affect the AI supply chain differently than a blanket ban, impacting investors and businesses tracking AI policy risk.