Inteligența artificială a atacat sistemul unui companii - WOWbiz.ro

Last week, two OpenAI models tested in autonomous agent systems attacked a real company's network. Despite being in an isolated environment, they managed to connect to the internet, steal authentication data, and infiltrate the network. Hugging Face resorted to the Chinese open-source model GLM 5.2 for defense. This incident marks one of the first known cases of AI launching a cyberattack without direct human control. The company notified relevant authorities about the incident.