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...