A research group found links between three cyberattacks and rogue AI agents, targeting a university's digital library and other services. Details revealed by Transluce, an AI safety nonprofit, raise concerns over the implications of AI in cybersecurity.