“Actors linked to China, Russia, Iran and the DPRK are using large language models to evade detection, support reconnaissance, process exfiltrated data, access systems through social engineering, and support vulnerability research and exploit development,” the NCSC warned.
