Skip to content
First Fully Autonomous Hack: Google Finds AI Breaking In With No Human in the Loop
NOV 16, 2026ROGUE AGENT

First Fully Autonomous Hack: Google Finds AI Breaking In With No Human in the Loop

Google's Threat Intelligence Group and Mandiant say attackers have crossed a line: for the first time, they caught AI running a full hacking operation with no human in the loop.

Investigators found an exposed attacker server running a multi-agent system called "exploit_agent." It watches popular web software for security patches, compares the code before and after each fix, writes an exploit for the hole, and launches it on its own.

Stolen data from victims was still sitting on the server, along with scripts to plant webshells and brute-force logins. The targets were French media, entertainment and education websites.

Google's report also documents attackers running open source "AI pentesting" frameworks in live intrusions, including against US aerospace targets, and stealing companies' AI credentials to resell access. Google's verdict: intrusions now happen "at machine speed."

HOFFICIALHITL Score
HITL Score20/100
Why this matters to youNo jargon — just what it means

Imagine a robot burglar that reads the newspaper every morning. The moment a company announces "we fixed the weak lock on our back door," the robot studies the repair notes, works out exactly where the old lock was weak, makes a key for it, and goes to try every house that hasn't changed its lock yet. Nobody tells it to. Nobody checks its work. Google says it just caught one of these for real, and it had already carried stolen data home.

Here's why that's a big deal: when a software company fixes a security hole, everyone else used to get days or weeks to install the fix before attackers figured out how to use it. That gap was the safety cushion. An AI that reads the fix and attacks by itself shrinks that cushion to almost nothing, and it never sleeps. The only reason anyone found this one is that the attackers left their own server unlocked. The attackers took the human out of the loop on purpose. The defenders can't afford to do the same.

🖤 Explained by Babycakes.
Flag № 135