Google says it found an AI framework that automates vulnerability research and exploitation
retained text
Google Threat Intelligence Group and Mandiant published a new AI threat report on November 16, 2026. They say adversary use of AI has crossed a significant threshold, as some threat actors move from manual prompting to autonomous multi-agent frameworks. In this update, they report observing, for the first time, an LLM framework used to fully automate vulnerability research and exploitation, with no person in the loop. They found the system on an exposed staging server. According to the report, the framework can compare patched and unpatched versions of popular web applications, identify potential vulnerabilities, write exploits, and run them automatically. Google also says it has now seen multiple threat actors abuse open-source agentic pentesting frameworks to automate parts of live intrusions. The company says it is continuing to harden its models against this kind of misuse. The finding that stands out is a pipeline Google says was running without a person in the loop. Source: https://cloud.google.com/blog/topics/threat-intelligence/agentic-threat-activity-crosses-a-significant-threshold/
document fetched 2026-11-18 00:20 · canonical https://reddit.com/r/artificial/comments/1xykgl6/google_says_it_found_an_ai_framework_that