ANTHROPIC has launched Claude Fable 5, a high-capacity AI model with enhanced safety measures for sensitive domains like cybersecurity. This is the first model of its kind available for widespread access, featuring built-in restrictions that revert to a less capable version, Claude Opus 4.8, when necessary. Despite this, early usage shows 95% success without fallback. The AI's safety protocols were thoroughly tested internally and via an external bug bounty program with no critical vulnerabilities found.
In related news, participating partners in Project Glasswing are being upgraded to Claude Mythos 5, expanding access to around 150 new organizations. Pricing has been set at $10 per million input tokens and $50 per million output tokens for both models.