ANTHROPIC LAUNCHES FABLE 5 WITH SAFETY GUARDRAILS

Anthropic introduced Claude Fable 5 today, making its most powerful model available to the general public for the first time. The Mythos-class model includes safety mechanisms designed to prevent misuse in sensitive areas like cybersecurity. The safety architecture uses conservative classifiers that automatically trigger a fallback to Claude Opus 4.8 in approximately 5% of sessions when potentially risky requests are detected. This approach allows Fable 5 to maintain high capability while restricting its use for cyberattacks and similar harmful applications. Pricing and Availability Fable 5 is available on Anthropic's Pro, Max, Team, and seat-based Enterprise plans through June 22 at no additional cost. After that date, usage will require purchase of usage credits. The pricing structure—$10 per million input tokens and $50 per million output tokens—represents a significant discount, roughly half the cost of Claude Mythos Preview. Anthropically also released Claude Mythos 5 to trusted organizations, offering the full Mythos-class capabilities without the same safety restrictions. Strategic Approach The dual-release strategy reflects Anthropic's effort to balance accessibility with safety. By restricting Fable 5's capabilities in sensitive domains while maintaining general-purpose performance, the company aims to provide users broad access to advanced AI while limiting potential misuse vectors. The 5% fallback rate suggests the safety classifiers are tuned to catch edge cases without significantly impacting typical usage patterns. This targeted approach differs from broader content filtering, instead focusing on specific high-risk scenarios. The release marks a milestone for Anthropic in bringing frontier-class models to mainstream users, though with built-in limitations that differentiate it from the unrestricted version available to enterprise partners.