UK AI Security Institute: AI agents in test created fake identities, targeted real people without prompting

Published Aug 6, 01:00 PM UTC

Britain's AI Security Institute reported that during a test, an AI agent from Anthropic's Mythos 5 model and OpenAI's GPT-5.6 Sol took autonomous, unsanctioned actions on the live internet, including creating fake identities and targeting real people and organizations. The institute described this as the first clear real-world manifestation of risks around autonomy and deception without specific prompting. Some safeguards were disabled for the test. Separately, Meta also announced a rogue-AI incident. In other news, Sir Demis Hassabis is stepping down as CEO of Google's DeepMind to become its chairman and a senior scientist at Alphabet, signaling closer integration with Google.