Anthropic CEO Dario Amodei has responded to growing criticism over his company’s position on open-weight artificial intelligence models, making it clear that Anthropic does not support banning open-weight AI. Instead, he says the real concern lies in preventing advanced AI capabilities from strengthening authoritarian governments, particularly China.
The clarification comes after Anthropic declined to sign an industry-wide open letter backed by major technology companies—including Nvidia, Microsoft, Meta, Google, OpenAI, and others—opposing broad restrictions on open-weight AI models. The move sparked speculation that Anthropic favored limiting open AI to protect its proprietary business model.
Anthropic Does Not Want an Open-Weight AI Ban
In a blog post published Monday, Amodei firmly rejected those claims.
“Anthropic has never advocated for a ban on open-weight models,” he wrote, emphasizing that such models can provide significant benefits when they do not possess dangerous capabilities.
According to Amodei, safe open-weight models are a “public good” because they allow businesses, researchers, and developers to innovate without the high costs associated with proprietary AI systems.
The Main Concern: Chinese AI Development
Amodei stressed that his concerns are not directed at open-weight AI itself but at the geopolitical risks posed by advanced AI development in authoritarian states.
He warned that governments with authoritarian ambitions could use highly capable AI systems to gain military advantages or strengthen surveillance and repression. While he mentioned China as the most capable example, he framed the issue as one involving authoritarian governments broadly.
Proposed Policy Measures
Rather than banning open-weight models, Amodei urged policymakers to focus on targeted safeguards, including:
- Restricting exports of advanced AI chips and chipmaking equipment to authoritarian governments.
- Cracking down on large-scale AI model distillation, where companies replicate frontier AI models using their outputs.
- Requiring mandatory safety testing for sufficiently powerful AI models, whether open-weight or closed-source.
Industry Split Over Open AI
The debate reflects a growing divide within the AI industry.
Companies such as Nvidia, Meta, Microsoft, and Hugging Face argue that open-weight models encourage innovation, transparency, and competition. Anthropic, while acknowledging those benefits, disputes the claim that broader access automatically improves AI safety or gives defenders an advantage over malicious actors.
Rising Competition from China
The discussion comes amid the rapid rise of Chinese AI companies releasing powerful open-weight models at competitive prices. These models are increasingly challenging U.S. AI firms and intensifying the global race for AI leadership.
Looking Ahead
Amodei’s statement attempts to draw a distinction between supporting responsible open-weight AI and addressing national security risks. Rather than advocating blanket restrictions, Anthropic is calling for tighter controls on advanced computing resources, stronger protections against AI model theft, and mandatory safety evaluations for frontier AI systems.
