From TokenPost · · 50s
Google AI Framework Automates Exploit Discovery
Google's AI system found in a staging server can automatically identify and exploit software vulnerabilities without human help.
In brief
Google's AI system found in a staging server can automatically identify and exploit software vulnerabilities without human help. Google discovered an AI framework configured to automate vulnerability discovery and exploitation on a staging server. Originally reported by TokenPost.
Google Threat Intelligence Group Finds AI Exploit System

Google Threat Intelligence Group and Mandiant discovered an exposed staging server hosting an AI framework configured to identify vulnerabilities through patch analysis and execute automated exploits.
AI Framework Uses LLMs for Vulnerability Scanning
The exploit_agent framework connected to frontier and locally run large language models to continuously analyze code changes around patch announcements for potential weaknesses.
No Exploit Successes Confirmed for AI System
The findings do not provide an exploit-run count, success rate, or specific vulnerabilities successfully exploited by the AI framework.
In short
Google discovered an AI framework configured to automate vulnerability discovery and exploitation on a staging server.
Read the TL;DR — this story on one card.





