Safety · The Decoder ·
New Claude Mythos becomes the first AI model to clear all cyberattack simulations from Britain's AI safety agency
The UK’s AI Security Institute says AI cyber capabilities are improving faster than previously estimated. Anthropic’s Claude Mythos Preview and OpenAI’s GPT-5.5 exceeded the institute’s revised timeline, and Mythos is the first model to pass all of AISI’s attack simulations.