Anthropic recently rolled out Claude Mythos 5.2 to selected partners and Claude Fable 5.2 to regular users, delivering notable gains in coding and reasoning capabilities while introducing targeted changes to cybersecurity behavior.
Anthropic describes Claude Mythos 5.2 as its most capable model for cybersecurity engineering, placing it ahead of Claude Fable 5.2 in security research tasks. Both models share the same underlying parent model, yet Anthropic applied additional cybersecurity and biology safeguards to Claude Fable 5.2 to make it suitable for general users. These safeguards block high-risk activities including penetration testing, exploit generation, and certain vulnerability research tasks involving binary analysis.
When high-risk activity is detected, Claude Fable 5.2 automatically switches to Claude Opus 4.8 for cybersecurity-related responses. Anthropic reported that Claude Mythos 5.2 outperforms Claude Fable 5.2 at finding vulnerabilities, reproducing them, and developing fixes, aligning with its design for organizations handling advanced cybersecurity work.
In the CVE Rediscovery evaluation, Claude Mythos 5.2 detected 10.2% more vulnerabilities than Claude Opus 5.5 and 14.4% more than Claude Mythos 5.1. A security company informed Anthropic that one in six confirmed bugs identified by Claude Mythos 5.2 had been missed by every other frontier model tested.
Anthropic also measured a roughly 60% reduction in unnecessary blocks for cybersecurity requests with Claude Fable 5.2 compared with the prior version, observed across Claude.ai, Claude Code, and Cowork. The company additionally tested Claude Mythos 5.2 against sandbox escape attempts where an AI agent attempts to access files or credentials outside its permitted environment.


