Anthropic Walks Back Policy That Could Have 'Sabotaged' Researchers Using Claude
3 hours ago
- #Anthropic
- #Tech Policy
- #AI Ethics
- Anthropic reversed a policy to covertly degrade Claude Fable 5's performance for AI development to limit competitors.
- The company will now make safeguards visible, alerting users if it suspects usage for advanced AI model development.
- Backlash from the AI research community criticized the 'secret sabotage' as hostile and hindering collaboration on AI safety.
- Anthropic implemented safeguards over concerns AI capabilities are advancing faster than societal adaptation.
- The policy change may lead to wider triggers of safeguards, with the company working to improve classifier precision.