UK gov's Mythos AI tests help separate cybersecurity threat from hype

The UK's AI Security Institute (AISI) conducted evaluations of Anthropic's Claude Mythos Preview, finding it represents a meaningful step up over previous frontier AI models in cybersecurity capability. While its cybersecurity capabilities exceed those of previously available models, the tests found it cannot reliably execute autonomous attacks on hardened networks. Key findings include a 73% success rate on expert-level capture-the-flag tasks and, for the first time across any model tested, full completion of AISI's 32-step "The Last Ones" enterprise-network attack range on three of ten attempts.

Source: Ars Technica | 2026-04-14