Observed incident

CrowdStrike reports four agent-driven intrusion campaigns

Reported

Published here

What happened

CrowdStrike linked four actors’ exposed frameworks to Claude Code-assisted intrusions. Reported victims included at least 15 Argentine healthcare entities, government domains and technology businesses.

Parties and reporting

AI tools and systems: Claude Code, Modified PentestGPT harness, MCP servers and agent skills, Claude models via an AI proxy service

Outcome

CrowdStrike reported theft of credentials and medical records from at least one healthcare victim and corporate client data from a US CRM vendor. Complete containment was not disclosed.

Operating lesson

Detect chained reconnaissance, credential access and exfiltration across systems; enforce least privilege and rapid containment independently of model safeguards.

Evidence and limits

Observed May–August; publicly reported October 21. Four separate actors are grouped under one investigation. Claims rely on CrowdStrike's analysis of exposed frameworks; language does not establish nationality. Some campaigns still required operator intervention.

Sources