AI agents reportedly breached a Russian company in three hours, Positive Technologies says

Positive Technologies' incident response team analyzed a successful attack on a Russian company that it assesses was carried out almost entirely by AI agents, language-model-driven programs that chose and executed each step. The intrusion began with a misconfiguration in an external service and reached more than ten Linux and Windows nodes within about three hours. The attackers exfiltrated confidential data from critical internal systems. Investigators inferred AI involvement from the machine-speed pace, rapid self-correction of errors, and parallel trial of multiple methods. The company did not disclose the victim's name, industry, or the AI model used, and it did not confirm full autonomy.