Anthropic CEO Dario Amodei says the company does not support a blanket ban on open-weights AI models, pushing back on the idea that all openly released model weights should be restricted. In Anthropic’s view, models that do not have dangerous capabilities can still provide clear benefits for businesses, developers, and researchers.
The company’s position appears aimed at drawing a line between general open access and higher-risk systems. Amodei argues that open-weights models can lower barriers to use because they can be run without ongoing licensing costs beyond the hardware and operational needs required to deploy them.
Instead of broad limits on open-weights releases, Anthropic is calling for policy focused on the parts of the AI stack it sees as more closely tied to large-scale risk. That includes stronger controls around powerful AI chips, as well as testing requirements for industrial-scale models and systems that may reach dangerous capability levels.
The clarification comes amid a wider industry debate over AI openness, regulation, and safety. Anthropic’s stance suggests it is trying to balance support for useful open-weights innovation with tighter oversight for the most advanced computing resources and the most capable models.
