Anthropic founder and CEO Dario Amodei clarified his company’s stance on open-weight AI models amid ongoing discussions in the tech industry. Speaking out against misinformation, Amodei emphasized that Anthropic has never supported banning open-weight models, distinguishing his position from concerns related to Chinese AI development. His comments followed an open letter from Nvidia and other AI firms urging policymakers to avoid premature restrictions on open-weight AI technologies.

Amodei views open-weight models without harmful capabilities as beneficial public goods, providing valuable resources for businesses, developers, and researchers at minimal cost beyond the computing power needed to run them. He made a clear distinction between his support for these models and his apprehensions about authoritarian regimes, particularly China’s Communist Party, deploying advanced AI to consolidate military advantage or suppress citizens. His unease centers more on national security and the potential misuse of AI rather than on the open-weight model ecosystem itself.

The Anthropic CEO highlighted the risks associated with open-weight models in scenarios like biological attacks, where regulating or monitoring such freely available AI becomes notably challenging. Citing a UK AI Security Institute report, Amodei warned that once open-weight models are released, they cannot be retracted, complicating safety efforts. He advocates for measures to curb Chinese access to powerful chips and to crack down on techniques like distillation, which are allegedly used to illicitly replicate foreign AI models.

Amodei also expressed strong support for international collaboration on AI safety testing, ideally involving all countries, including China. He endorsed proposals for a global framework to evaluate frontier AI systems regardless of origin, suggesting this approach could help prevent the weaponization of AI and limit biological threats. His vision underscores the need for cooperation even in geopolitically tense environments to ensure AI benefits are widely shared and its risks contained.