Claude Mythos only model to complete full cyber kill chain, experts say

Booz Allen’s Cyber Weapon Index evaluated 18 US and Chinese AI models for autonomous offensive cyber capability and found Anthropic’s Claude Mythos uniquely able to complete a full kill chain and exploit a real vulnerability; several other models achieved domain-level access or lateral movement. The report ranks models by combined vulnerability-research and kill-chain scores, highlights the amplifying role of attack harnesses, warns that many models could reach Mythos-level weaponization within months, and urges U.S. action to harden critical infrastructure and develop AI-enabled offense/defense “overmatch.”