Anthropic reverses policy that could block researchers using Claude
A controversial clause threatened to limit Claude's use for AI safety research.
Deep Dive
Key Points
- Anthropic's original policy banned using Claude to 'improve or train any competing AI model'—a clause many researchers said would silence safety work.
- After public outcry, Anthropic walked back the language, calling it 'poorly scoped' and reaffirming support for open research.
- The reversal ensures Claude remains available for non-commercial AI safety research, though direct commercial competition is still prohibited.
Why It Matters
Policing research on safety models could stifle independent scrutiny—this reversal protects open AI evaluation.