Mythos & GPT-5.5: Independent Evals Confirm Frontier AI Is Now Competent at Full Cyberattack Chains

Two independent security evaluations XBOW and the UK AI Security Institute confirm that frontier AI models can now execute full end-to-end cyberattack chains. Mythos scored 68.6% and GPT-5.5 scored 71.4% on expert cyber tasks. The window to prepare is now.

By Signal.lab Editorial | Editorial at Signal.lab

Canonical: https://signal-lab.connxr.com/insights/mythos-gpt-55-cyber-capability-curve