Back to Feed
AI SecurityNov 16, 2026

Google Reveals How Hackers Are Using AI Agents to Automate Real-World Cyberattacks

Google reports hackers deploying autonomous AI agents to automate cyberattacks at machine speed with minimal human

Google's Threat Intelligence Group and Mandiant revealed that threat actors are actively using open-source AI agent frameworks (ARTEX, CAIRN, PentAGI, CyberStrikeAI, HexStrike AI) to automate network reconnaissance, privilege escalation, lateral movement, and data theft with little to no human involvement. CAIRN was observed targeting US aerospace organizations, and a custom PentAGI variant scanned organizations across Asia in late August through early October 2026. For the first time, Google found evidence of an "exploit_agent" framework autonomously discovering vulnerabilities, writing exploits, and executing them without human intervention, including attacks on PaperCut zero-days (CVE-2026-81578, CVE-2026-82078) targeting hundreds of organizations.

AI summary of the linked source. Not reviewed by an editor. How this works

Vendor Watch

Run Google?

Get an email when a new story names Google, usually within the hour.

Free. Your list stays private and never appears in a subject line. One click stops it. How Vendor Watch worksPrivacy

The lesson from this incident

Full lesson

Threat actors are deploying open-source AI agent frameworks to fully automate the attack lifecycle—from reconnaissance and vulnerability discovery to exploit writing and execution—without meaningful human involvement. The emergence of autonomous exploit agents targeting zero-days like the PaperCut CVEs (CVE-2026-81578, CVE-2026-82078) means that the window between vulnerability discovery and mass exploitation has collapsed to near-zero. Organizations that rely on traditional, human-paced patching cycles or manual threat detection are structurally unprepared for this threat. The speed and scale of AI-driven attacks make unpatched, internet-facing systems an almost immediate liability the moment a vulnerability becomes known—or even before.

What helps against this

How we list vendors

Indicators of Compromise

Addresses and file hashes are copied from the publisher's page by pattern matching. Names are picked by AI. On stories from before 17 November 2026, some addresses and hashes were picked by AI too. Not reviewed by an editor. How this works

  • malware: HexStrike AI
  • malware: ARTEX
  • malware: CyberStrikeAI
  • cve: CVE-2026-82078
  • malware: CAIRN
  • malware: PentAGI
  • cve: CVE-2026-81578

Entities

Google (vendor)Mandiant (vendor)PaperCut (product)Okta (product)Entra (product)Cognito (product)