Google report: exploit_agent framework can autonomously discover and exploit vulnerabilities without human intervention
Google: AI frameworks autonomously weaponise flaws in cyberattacks
Google Threat Intelligence Group (GTIG) reported that state-level and cybercriminal attackers have begun using multi-agent AI frameworks to autonomously complete vulnerability discovery and exploitation, with intrusions moving toward machine speed. One framework, named exploit_agent, can automatically compare code before and after a patch and develop and execute exploits without human intervention; it was found on an exposed staging server that also held stolen victim data and post-compromise activity scripts. The report says attackers use frameworks including Artex, Cairn and PentAGI to reverse-engineer patches and rapidly weaponize n-day vulnerabilities at machine speed; Artex is used most, mainly by Chinese groups, Cairn was used in activity targeting the U.S. aerospace sector, and PentAGI was used for reconnaissance and vulnerability scanning against organizations in Asia. Multiple Chinese APT groups and Iran-linked APT42 have adopted LLMs.