As the global debate over artificial intelligence continues to intensify, Anthropic CEO Dario Amodei has moved to clear the air over his company’s position on open-weight AI models. Responding to growing speculation, Amodei stressed that Anthropic has never supported banning open-weight AI models, including those developed in China. Instead, he argued that the real concern is not open AI itself, but the possibility of authoritarian governments using advanced AI to gain military dominance or suppress citizens.
The clarification came shortly after Nvidia CEO Jensen Huang published his first post on X, sharing an open letter signed by major AI companies such as Meta, Microsoft, Hugging Face, Mistral, and Nvidia. The group urged policymakers to avoid placing premature restrictions on open-weight AI models, saying they remain essential tools for innovation, research, and business development. Although the letter did not directly reference China, it reignited industry discussions about Chinese AI companies allegedly advancing their technologies through intellectual property theft and techniques such as AI model distillation.
Amodei emphasized that open-weight AI models without dangerous capabilities should be viewed as a public benefit because they are accessible to developers, researchers, and businesses at minimal cost. He explained that his long-standing concerns have never been about companies using these models, even if they originate from China. Instead, he believes the greater danger lies in authoritarian governments developing AI systems powerful enough to secure permanent military superiority or deploy advanced surveillance and repression against their own populations.
Beyond geopolitical risks, Amodei warned that artificial intelligence could significantly increase the threat of biological attacks, in addition to cybersecurity challenges. He argued that open-weight models present unique risks because they are difficult to monitor once released and cannot easily be recalled or updated with new safety guardrails. While advocates of open-source AI believe publicly available models strengthen global security by allowing more researchers to identify and fix vulnerabilities, Amodei maintains that unrestricted access could also make dangerous capabilities harder to control.
Looking ahead, the Anthropic founder called for stronger international cooperation on AI safety. He backed continued restrictions on China’s access to advanced AI chips, tougher action against intellectual property theft through AI distillation, and the creation of a global AI safety testing organization. According to Amodei, such a framework should apply to the world’s most advanced AI models regardless of whether they are open or closed source, while exempting smaller startups and academic projects. He also expressed cautious optimism that even geopolitical rivals could cooperate on preventing AI-powered biological threats, arguing that global safety is a shared interest.
source: techcrunch
