Google GTIG finds attackers using open source multi-agent AI frameworks in live intrusions
GTIG says it saw a threat actor fully automate vulnerability discovery and exploitation with no human in the loop for the first time, and zero-day exploits hit 22 in August 2026.
What happened
Google's threat team says hackers now use AI agents, which are programs that act on their own, to break into computers. Some tools handle many steps of an attack with little human input. Google also found one system that read software patches, wrote exploits and ran them automatically.
Why it matters to you
Attackers can move faster from a published software fix to a working attack. Companies have less time to install updates before someone exploits the flaw.
The takeaway
Patch known flaws quickly, because AI tools shorten the time attackers need to exploit them.
What it means for you
Patch diffing and exploit generation can now be automated, so shrink the time between a vendor patch and your deployment.
Internet-facing servers with unpatched known flaws are the main target, and GTIG says the patch window is shrinking.
Updates
Label changed from Reported to Confirmed after cyberscoop.com reported independently.
Added source: cyberscoop.com.
Newest first. Labels change only by the rules in our standards.