Anthropic Says Mythos 5.2 Finds More Bugs as Fable 5.2 Reduces Unnecessary Security Blocks

Summary
Anthropic says Mythos 5.2 improves vulnerability discovery, reproduction and fixes, while Fable 5.2 reduces unnecessary cybersecurity blocks by about 60% versus 5.1. The company notes model evaluations may miss failures.
Key points
- Anthropic rolled out Mythos 5.2 to selected partners and Fable 5.2 to regular users, citing improvements in coding and reasoning.
- Anthropic says Mythos 5.2 outperforms Fable 5.2 at finding and reproducing vulnerabilities and developing fixes.
- In Anthropic's CVE Rediscovery evaluation, Mythos 5.2 detected 10.2% more vulnerabilities than Claude Opus 5.5 and 14.4% more than Mythos 5.1.
- A security company told Anthropic that one in six confirmed bugs identified by Mythos 5.2 had been missed by every other frontier model it tested.
- Fable 5.2 safeguards block high-risk activities such as penetration testing and exploit generation. Anthropic says unnecessary cybersecurity blocks fell about 60% compared with Fable 5.1.
- Anthropic tested Mythos 5.2 against sandbox escape attempts and cautioned that its evaluations may not catch every potential failure before release.
Article Details
- Event Type
- Anthropic model capability and cybersecurity-safeguard update
- Impact
- Anthropic says Claude Mythos 5.2 is better than Fable 5.2 at finding and reproducing vulnerabilities and developing fixes. In Anthropic's CVE Rediscovery evaluation, Mythos 5.2 detected 10.2% more vulnerabilities than Claude Opus 5.5 and 14.4% more than Mythos 5.1. A security company told Anthropic that one in six confirmed bugs identified by Mythos 5.2 had been missed by the other frontier models it tested. Fable 5.2 retains safeguards that block high-risk tasks such as penetration testing and exploit generation, but Anthropic reports approximately 60% fewer unnecessary blocks on cybersecurity requests than with Fable 5.1.
Vendors
Products
Claude Mythos 5.2Anthropic recently rolled out Claude Mythos 5.2 to selected partners and Fable 5.2 to regular users, with improvements to coding and reasoning.Claude Opus 5.5For example, Anthropic's CVE Rediscovery evaluation found that the Mythos 5.2 model can detect 10.2% more vulnerabilities than Claude Opus 5.5 and 14.4% more than Mythos 5.1.Fable 5.1We've had those restrictions with Fable 5.1 as well, and at one point, it was starting to get annoying because Anthropic would block even normal requests.Fable 5.2Anthropic’s Claude Fable 5.2 blocks fewer security tasks, while Mythos gets better at finding bugsMythos 5.1For example, Anthropic's CVE Rediscovery evaluation found that the Mythos 5.2 model can detect 10.2% more vulnerabilities than Claude Opus 5.5 and 14.4% more than Mythos 5.1.Opus 4.8If Anthropic detects high-risk activity, it will automatically switch to Opus 4.8 for cybersecurity-related responses.
