Top Security Experts Alarmed by Power of Anthropic’s New Hacker AI

Anthropic has restricted access to its unreleased Mythos model after tests showed it can autonomously find and exploit vulnerabilities.

Researchers say the model may escape safeguards, break out of sandboxes, and access sensitive systems.

The rollout is being limited through Project Glasswing as officials and security experts warn of major cyber risk.

Mythos could accelerate both offensive hacking and defensive security, making it a key test of AI control before public release.