Amodei Rejects Open-Weights Ban, Backs Global AI Tests
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Dario Amodei published a position paper saying Anthropic has never backed an open-weights model ban and calling for global safety testing of frontier models
- Amodei listed reasons top chips shouldn't be sold to China, per Anthropic's statement, framing restrictions as a national-security measure rather than an anti-open-weights stance
- OpenAI and Google are backing open weights, per The New Stack's headline, directly contrasting with Anthropic's call for testing requirements
- The Wall Street Journal opinion piece and The Register headline pushed back on the testing framework, with Register framing opposition to safety requirements as 'thumb on the scales against open-weights fearmongering'
- David Sacks, posting on X, called Anthropic's stance hypocritical, citing its position on training data and IP as inconsistent with how it views competitors building on its output
- Politico distilled Amodei's position as 'Don't ban cheap AI — but clamp down on China,' highlighting the bifurcated stance separating open-weight development from Chinese access to compute
Why it matters: Amodei's middle path splits the AI industry: OpenAI and Google back open development, while the WSJ opinion page and David Sacks publicly challenge the safety-test framework as a 'permission slip.' The bifurcation leaves frontier-model testing as the most politically viable compromise but fractures the coalition that might otherwise oppose it, while keeping chip-export restrictions targeting China on the table as a separate policy track.
Ask SkimNews
