Anthropic’s Claude Mythos Preview shocks with offensive cyber capabilities

On April 7, 2026, Anthropic released the Claude Mythos Preview, a frontier AI model that demonstrated exceptional cyberattack abilities, including detecting a 27-year-old Linux vulnerability and generating exploit code. The UK AI Security Institute (AISI) assessed it using the TLO framework, finding it outperformed previous models, completing all 32 stages in 3 out of 10 attempts. The model reportedly analyzed a Windows OS vulnerability and created an attack tool in just 31 minutes.