Google warns fully automated cyberattacks have arrived and will worsen quickly
Google's Threat Intelligence Group has reported the first cases it has seen in which attackers find and exploit security holes with no person in the loop at all. The group documented this on an exposed server running an AI framework named exploit_agent, which it describes in the GTIG AI Threat Tracker. The software connects to several language models, compares program code from before and after a patch, spots the underlying weakness, writes an exploit, and launches it on its own.
Comments in Russian in the configuration files point to a Russian-speaking operator. Data on the same server indicate the campaign was aimed at French sites in media, entertainment, and education. GTIG treats the move to fully automated discovery and exploitation as a critical turning point and expects the pattern to grow more dangerous quickly. For now, it says, these attacks trade stealth for speed and leave forensic traces that make them easier to detect than quieter, human-driven operations.
Continue in the app — vote & join in ➔