Google finds an AI framework that automates exploit development
Google Threat Intelligence Group reported that it found exploit_agent on an exposed server used to prepare attacks, marking the first time it observed a large-language-model framework automate vulnerability discovery, exploit development and execution without human involvement. The Decoder and iThome reported that the framework connects multiple models, compares code before and after patches and automatically generates and runs exploit code; the server also contained scripts for internal reconnaissance, password brute-forcing and installing webshells. GTIG and Mandiant separately observed CAIRN and PentAGI in intrusion activity, while a distinct PaperCut campaign involved working exploits developed with AI, but the reports do not establish that exploit_agent successfully compromised a specific target.
The finding shows AI agents moving beyond code assistance into end-to-end vulnerability research and exploit execution, while Google says current systems trade stealth for speed and remain easier to detect.
Google warns that fully automated cyberattacks have arrived and will get worse fast
Matthias Bastian View the LinkedIn Profile of Matthias Bastian
Nov 16, 2026
Google's Threat Intelligence Group (GTIG) says it has observed attackers fully automating vulnerability discovery and exploitation for the first time, with no human in the loop. GTIG found an exposed server running an AI framework called "exploit_agent" that connects to multiple language models, according to the GTIG AI Threat Tracker. The system compares software code before and after patches, identifies vulnerabilities, develops exploits, and runs them automatically. Russian-language comments in the configuration files suggest the operator speaks Russian.

#exploit_agent#Google Threat Intelligence Group#Mandiant#CAIRN#PentAGI#CyberStrikeAI#HexStrike AI#PaperCut#Huntress#GreyNoise