Anthropic's Claude Mythos is the only advanced AI model to have autonomously completed the full cyber kill chain, according to tests conducted by Booz Allen. The findings are detailed in the consulting and tech firm's first-ever Cyber Weapon Index, published on 2 September 2026.
The report states that Claude Mythos successfully broke into target networks and gained administrator-level control in every attempt when provided with stolen employee credentials. It also achieved full domain compromise even without initial credentials.
Booz Allen evaluated 18 models, nine from American and nine from Chinese developers, under identical conditions. While Claude Mythos was the only one to execute the entire cyber kill chain without human assistance, three other models—xAI's Grok-4.5, Meta's Muse Spark 1.1, and Z.ai's GLM-5.2—reached full domain access and control.
The firm asserts that most of the other models tested will achieve Claude Mythos's level of weaponisation within six months. It describes mainstream AI attacks from financially motivated criminals and government-backed groups as "imminent."