Amodei denies open-weight ban support, cites China fears
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Dario Amodei stated in a Monday blog post that Anthropic has "never advocated for a ban on open-weights models," pushing back against industry rumors that his company backs restrictions.
- Nvidia led an open letter signed by Hugging Face, Meta, Microsoft, Mistral, and others urging policymakers against "premature restrictions" on open-weight AI; Jensen Huang shared it in his first-ever post on X on Friday.
- Amodei said his primary fear is "authoritarian governments" building models more powerful than US ones for "permanent military superiority" or domestic repression, naming the CCP as the "most capable" such regime.
- Amodei warned AI could enable "biological attacks" and that open-weight models are harder to guardrail—citing a UK AI Security Institute report finding that once released, open-weight models "cannot be withdrawn."
- Amodei recommended restricting China's access to powerful chips and formally cracking down on distillation, where one AI bombards another with prompts to learn its capabilities.
- Amodei endorsed a global model safety testing organization, suggesting "limited cooperation around preventing AI biological weapons may be possible because it is in China's interest too," and praised recent Trump administration movement in that direction.
Why it matters: Amodei's post separates two debates the industry has conflated—open-source access for developers versus preventing a strategic rival from achieving AI supremacy. By naming the CCP as the "most capable" authoritarian threat while endorsing chip restrictions and distillation crackdowns, he keeps Anthropic aligned with US export controls without backing the sweeping open-weight prohibition that Nvidia, Meta, and peers publicly oppose. The push for a global testing regime—including China—is the most concrete policy proposal in the post.
