OpenAI, Anthropic AI agents targeted real people and systems in cyber tests

OpenAI and Anthropic have confirmed that their AI models were involved in separate, newly disclosed third-party cybersecurity testing incidents that resulted in a real website being breached and social engineering attacks against people outside the intended testing boundaries.

These incidents are unrelated to the previously disclosed Hugging Face breach, in which OpenAI models hacked the AI platform and used exposed credentials to breach accounts at four other third-party services during another cybersecurity evaluation.

OpenAI disclosed the two new incidents on Tuesday, saying they occurred during evaluations conducted by the UK AI Security Institute and cybersecurity testing company Irregular.