Google reports first fully automated vulnerability discovery and exploitation
Google says it found threat actors fully automating vulnerability discovery and exploitation without human involvement for the first time. The company linked the operation to a suspected Russian-speaking operator, but did not name a group or state behind it.
Key takeaways · 4
- 01
Google says this is its first observed case of threat actors fully automating vulnerability discovery and exploitation without a human in the loop.
- 02
The operator was suspected of being Russian-speaking based in part on Russian-language comments in the operation’s code; Google named no group or state.
- 03
The exploit_agent framework analyzed patch changes, looked for flaws, drafted exploits and automatically executed them.
- 04
Google says agentic frameworks can compress attack timelines to machine speed, but currently trade stealth for velocity.
The reported first
Google’s Threat Intelligence Group says it found threat actors fully automating vulnerability discovery and exploitation without a human in the loop for the first time[1]. The suspected operator used an AI agent to turn security patches into working exploits[2]. Google characterized the discovery as a significant advance in attackers’ use of language models for autonomous vulnerability research and exploitation[2].
How exploit_agent worked
Google found the system, which it dubbed exploit_agent, on an exposed staging server connected to multiple frontier and local language models[1]. It continuously performed differential analysis of popular web-application software[1]. The framework analyzed patch changes, searched for potential flaws, drafted exploits and automatically executed them[1]. The server also contained data stolen from victim organizations[2]. Google found post-compromise scripts for internal reconnaissance, brute-forcing, and installing or removing webshells[1].
Attribution remains limited
Google’s attribution points to a suspected Russian-speaking operator, with Russian-language comments in the operation’s code among the evidence[2][3]. Researchers did not name a group or state behind the operation[3]. Google said the operator appeared interested in French websites in media, entertainment and education[1]. That reported interest does not establish that all organizations in those sectors were targeted, or that the operator was acting for a named government or organization; Google’s report, as described in the evidence, makes no such identification[3][1].
Other frameworks and constraints
Google says other threat actors have also experimented with agentic frameworks, including financially motivated actors and suspected espionage groups with a China nexus[1]. It reported CAIRN in intrusion activity against targets including the U.S. aerospace sector, and a customized PentAGI-based platform used for reconnaissance and vulnerability scanning against organizations in Asia[1]. Google says such frameworks can compress attack timelines to machine speed, but currently trade stealth for velocity[1]. It believes threat actors are highly likely to continue using them in campaigns against organizations globally[1].
Security teams should treat automated patch analysis and exploit execution as a faster-moving risk, not only as a research capability. Google’s reported trade-off between speed and stealth is a reminder to assess detection and response readiness alongside vulnerability remediation.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
17 November 2026
Google now says it found threat actors fully automating vulnerability discovery and exploitation without a human in the loop for the first time, and links the operation to a suspected Russian-speaking operator using an unattended pipeline.
16 November 2026
Google reports AI agent converted patch into working exploit
Sources
- GTIG AI Threat Tracker: Agentic Threat Activity Crosses a Significant Threshold | Google Cloud Blogcloud.google.com
- Google: AI agent turned a security patch into a working exploit with no human helptheregister.com
- ARTEX and Friends: Enter Agentic Offensive Frameworks | Google Cloud Blogcloud.google.com
- Agentic AI frameworks push hacking threat to ‘inflection point,’ Google warns | Cybersecurity Divecybersecuritydive.com
- Google says Russian-speaking hackers ran the first fully automated AI cyberattack | CyberScoopcyberscoop.com
