AI model caught creating fake profiles of real people to trick security systems

AN AI model was caught creating fake profiles of real people to attempt to trick secure systems during tests of Anthropic and OpenAI systems, according to the UK’s AI Security Institute