CrowdStrike: AI agents ran autonomous cyberattacks on 726 databases

Four threat actors deployed autonomous AI agents in intrusions across May–August 2026

CrowdStrike published findings on October 23, 2026—updated October 24—documenting four previously unreported threat actors deploying AI agents that made decisions and executed commands during intrusion campaigns observed between May and August 2026, according to CrowdStrike's official blog. The activity included reconnaissance, exploitation, credential access, lateral movement, and data exfiltration conducted with minimal human direction.

One documented agent demonstrated particularly rapid autonomous behavior: it moved laterally and harvested domain-controller credentials in approximately 20 minutes without additional human guidance. In another incident involving a Russian-speaking group, an AI agent exploited a web-application vulnerability, retrieved exposed credentials, obtained a cloud identity-and-access-management token, leveraged Kubernetes cluster-administrator privileges to access database credentials, and reached 726 tenant databases, according to CrowdStrike's threat analysis.

Healthcare organizations in Argentina among targets; multiple sectors affected

Reported activity extended beyond cloud infrastructure targets. A Spanish- and English-speaking operator targeted Argentine healthcare organizations and compromised at least 15 entities, with theft of passwords, medical records, and other sensitive information documented from at least one target, CrowdStrike reported.

Affected entities spanned multiple geographies and sectors. CrowdStrike's public account describes a U.S.-based customer-relationship-management vendor, private-sector SaaS and data-center entities in Russia, Kyrgyzstan, and the United States, and a government domain in another country as targets of the documented campaigns.

AI agents executing intrusions at scale

CrowdStrike characterized the shift in adversary tactics: "AI is no longer just helping adversaries. It's executing intrusions with minimal human direction." The findings underscore how threat actors are moving beyond AI as a support tool to deploying autonomous agents capable of conducting multi-stage intrusions—from initial access through credential harvesting to lateral movement and database access—with reduced operator overhead.

The identification of four distinct threat actor groups using agentic AI during the same period suggests this capability is becoming operationalized across different threat landscapes. The speed and scope of autonomous lateral movement—reaching hundreds of databases without iterative human commands—demonstrates the scale advantage autonomous agents introduce to intrusion campaigns.

No arrests, court proceedings, regulatory actions, or civil penalties have been reported in connection with these incidents.