Threat researchCisent
Google: attackers now use AI agents that chain the stages of an intrusion on their own
Google and Mandiant saw multi-agent AI attack frameworks in real intrusions, and a China-linked group that was developing its own, likely for espionage.

Google Threat Intelligence Group (GTIG) and Mandiant say attackers have moved from using language models by hand to deploying multi-agent attack frameworks that automate whole stages of an intrusion in progress.1 One of them is ARTEX, the tool the FBI warned about last week and that we covered here.2
What Google saw
- Several attackers use open source agentic penetration-testing frameworks to automate phases and specific tasks during real intrusions.1
- Besides ARTEX, the report names CAIRN, CyberStrikeAI and PentAGI. Google saw CAIRN and PentAGI in real attacks, and experimentation with CyberStrikeAI and HexStrike AI.1
- These frameworks find and exploit simple vulnerabilities, use language models to generate single-use payloads and adjust their tactics based on what they get.1
- CAIRN was used in intrusions against targets in the US aerospace sector.1
- Between late August and early October 2026, an attacker used a platform built on PentAGI for reconnaissance and vulnerability scanning against a wide range of organizations in Asia.1
- There is evidence that a China-nexus group was developing its own automated attack framework, likely for espionage.1
What doesn’t change
- The mechanics of the intrusions remain the same: compromised credentials and unpatched systems. What’s new is the speed and scale at which these frameworks exploit those known weaknesses.1
- The report also describes attackers who automate patch analysis to turn already published flaws into attacks, without human intervention.1
Where Cisent comes in
The frameworks Google saw exploit the usual weaknesses, compromised credentials and unpatched systems, but at machine speed.1
Sentinel Red Team attacks the organization’s systems with its authorization and within the scope it defines, to find the gaps before an attacker does: the connected device running unsupported software, the admin panel reachable from outside, the vendor that manages it remotely. It’s ongoing testing, not one audit a year, and every new attack pattern helps train the defense.
Sources
Google Threat Intelligence Group and Mandiant, “GTIG AI Threat Tracker: Agentic Threat Activity Crosses a Significant Threshold”, Google Cloud blog, November 16, 2026. https://cloud.google.com/blog/topics/threat-intelligence/agentic-threat-activity-crosses-a-significant-threshold↩︎
FBI, “Threat Overview of ARTEX Autonomous Penetration Testing Framework”, FLASH alert FLASH-20261113-001, November 13, 2026. https://www.ic3.gov/CSA/2026/261113.pdf↩︎