Anthropic Says Claude Fable 5.2 Had About 55% Fewer Safety Blocks

Anthropic said early testing found Claude Fable 5.2 triggered about 55% fewer safety blocks than version 5.1, while Claude Mythos 5.2 used outside files or credentials in 3.1% of sandbox-escape scenarios.

The reductions varied by category: cybersecurity-related blocks dropped about 60%, and biology-related blocks about 40%.

Anthropic also reported that Mythos 5.2 identified 89.6% of known flaws in the CVE Rediscovery test. In a separate sandbox-escape test, it used outside files or credentials in 3.1% of scenarios, compared with 39% for Mythos 5 and 21% for Mythos 5.1.

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 Monday, Nov. 9, using the same underlying system with different safeguards. Fable is available to all users; Mythos access requires approval through the Cyber Verification Program or Life Sciences Verification Program, continuing Anthropic’s split between broad Fable access and controlled Mythos use.

Claude Enterprise customers can use Claude Security, which runs on Mythos 5.2 to scan code for flaws and track findings until they are fixed.