Anthropic CEO Dario Amodei has clarified that the company does not support banning open-weight AI models, while drawing a clear line between open access and the risks he associates with advanced AI development in China.
In a Monday post, Amodei said Anthropic has never advocated for restrictions on open-weights, arguing that models without dangerous capabilities can serve as a public good for developers, businesses, and researchers. His comments followed a broader industry push, led by figures including Jensen Huang, to avoid sweeping limits on open-weight systems.
The debate has intensified around concerns that some Chinese AI labs may be advancing through distillation, a technique that can help one model learn from another. Amodei separated that issue from the value of open-weight tools, saying his main concern is the possibility that authoritarian governments could use frontier AI to gain strategic advantage or expand surveillance.
He also pointed to risks beyond cybersecurity, including the potential misuse of AI in biological contexts. In his view, once open-weights are released, they are difficult to fully control or retract, which makes safety governance especially important.
Amodei said he supports stronger chip controls, action against distillation, and broader model safety testing frameworks. He added that a global testing system, if widely adopted, could help create shared standards for advanced AI. That approach reflects a growing effort to balance openness, innovation, and responsible oversight as AI moves closer to everyday infrastructure.
