Anthropic Releases Claude Fable 5, Its Most Powerful AI Yet, With Cyber Safeguards

Anthropic released Fable 5 and a cyber-enabled twin, Mythos 5; classifiers on Fable route flagged cyber, biology, chemistry, and model-distillation requests to a weaker fallback, while Mythos 5 remains available to vetted defenders. Anthropic's red-team testing and Project Glasswing showed the model could autonomously find and produce working exploits (including zero-day and N-day exploits) across major OSes and browsers, enabling rapid exploit development that outpaces patching; this creates a defensive bottleneck and significant attacker uplift if similar models are widely available. The company also requires 30-day retention of traffic to Mythos-class models for safety monitoring and plans trusted-access programs for defenders.