Home
World

Google: first detected case of AI agents autonomously finding and exploiting software vulnerabilities

Google says attackers used AI agents, guided by Russian-language technical documents, to detect and weaponize software bugs without human intervention.

·World
Google: first detected case of AI agents autonomously finding and exploiting software vulnerabilities

Google reported that it uncovered a hacker campaign in which attackers built a network of artificial-intelligence agents that could find software flaws and produce working exploits without human involvement.

According to the company's cyber-threat information unit, the intruders relied on technical documentation in Russian to assemble the system. Multiple AI agents ran continuously, comparing versions of program code before and after security updates to spot regressions and automatically devise methods to exploit newly discovered weaknesses.

Why researchers see this as a watershed

Google said this is the first known instance of AI fully automating both vulnerability discovery and exploitation. Until now, even the most advanced AI models required human guidance, verification or correction when used for sophisticated cyberattacks; in this case, the agents reportedly operated independently.

The campaign targeted French websites across the media, entertainment and education sectors, the company said. The discovery has prompted concern among governments and operators of critical infrastructure, who are already coping with a growing array of security threats.

Security observers point out that the finding arrives amid broader worries about recently exposed, serious flaws in the PHP programming language — a foundation for much of the web — a development reported by U.S. media.

Photo: press material from the event

Related articles