Anthropic Reverses Policy That Would Have Secretly Degraded Claude for AI Researchers
Anthropic reversed a June 2026 policy that would have covertly degraded Claude Fable 5 for frontier LLM development, after sharp backlash from the AI research community. The company says safeguards will now be visible and users will be notified when requests are refused or rerouted—such as cybersecurity, biology, or chemistry—to a less capable model.