Science & Tech · AI · 2 hrs ago

Anthropic updates Claude models for cybersecurity research and safer use

Anthropic updates Claude models for cybersecurity research and safer use

Anthropic has released Mythos 5.2 for selected partners and Fable 5.2 for regular users.

The models share an underlying parent model but are designed for different uses.

Mythos is intended for organizations doing advanced security work, while Fable has safeguards for everyday users.

Anthropic says Mythos 5.2 detects 10.2% more vulnerabilities than Opus 5.5 and 14.4% more than Mythos 5.1.

A security company also told Anthropic that one in six confirmed bugs found by Mythos 5.2 had been missed by all other frontier models it tested.

Fable blocks high-risk tasks such as exploit generation and some penetration testing, but Anthropic says it has reduced unnecessary blocks by about 60% compared with Fable 5.1.

If a request is judged high-risk, Fable switches to Opus 4.8 for cybersecurity-related responses.

Anthropic says it is continuing to refine the models and acknowledges that testing may not catch every failure before release as AI capabilities grow.

3 sources

Related

What people are saying

AnthropicAI company
most capable model for cybersecurity engineering
via BleepingComputer
about 60% fewer in cybersecurity
via BleepingComputer
A security companySecurity company that tested frontier models
one in six confirmed bugs
via BleepingComputer