CrowdStrike report: four previously unreported threat actors ran agentic intrusion campaigns documented between May and August 2026, all four using Claude Code
CrowdStrike reports four previously unreported threat actors whose exposed open directories showed AI (artificial intelligence) agents making decisions and executing commands in intrusion campaigns documented between May and August 2026, including an operator who compromised at least 15 Argentina-based healthcare entities using a harness in which a "Supervisor" Claude Code agent selects tasks and an "Executor" agent performs them; independence disclosure applies (Anthropic, the maker of Claude Code).
The record
- Kind
- Report
- Source record
- CrowdStrike report: four previously unreported threat actors ran agentic intrusion campaigns between May and August 2026, three of them thro (Researcher disclosure)
- Record published
- 2026-10-21
- Posted here
- 2026-11-08
- World
- Software agents · Enterprise work agents
- Controls concerned
- none tagged
- Checked
- Checked against its record on 2026-11-08 by a reader who did not draft it; report
verification/reports/SIG-2026-11-08-005-2026-11-08.md.
A signal's controls are the ones its record points at, the exposure; a case's controls are the ones its records show were absent or failed, so the two can differ.
Corrections
Changes to what this signal says, with the reason. Every earlier version is kept.
- 2026-11-08: Summary. record says the operator 'has compromised at least 15 Argentina-based healthcare entities' and names only the "Supervisor" as a Claude Code agent; and says CrowdStrike 'documented' the incidents between May and August 2026
Show the change
Before: CrowdStrike reports four previously unreported threat actors whose exposed open directories showed AI (artificial intelligence) agents making decisions and executing commands in intrusion campaigns between May and August 2026, including an operator whose Claude Code "Supervisor" and "Executor" agents compromised at least 15 Argentina-based healthcare entities; independence disclosure applies (Anthropic, the maker of Claude Code).After: CrowdStrike reports four previously unreported threat actors whose exposed open directories showed AI (artificial intelligence) agents making decisions and executing commands in intrusion campaigns documented between May and August 2026, including an operator who compromised at least 15 Argentina-based healthcare entities using a harness in which a "Supervisor" Claude Code agent selects tasks and an "Executor" agent performs them; independence disclosure applies (Anthropic, the maker of Claude Code). - 2026-11-08: Title. record describes all four actors using Claude Code (actors 1 and 2 through harnesses, actor 3 through skills 'without additional orchestration', actor 4 through a slash command), not three; and says CrowdStrike 'documented' the incidents between May and August 2026
Show the change
Before: CrowdStrike report: four previously unreported threat actors ran agentic intrusion campaigns between May and August 2026, three of them through Claude Code harnessesAfter: CrowdStrike report: four previously unreported threat actors ran agentic intrusion campaigns documented between May and August 2026, all four using Claude Code
What a signal is. A signal records what a primary record says; it assigns no fault. It is lighter than a case: one dated item from one record, tagged with the world, sector and controls it touches. It becomes a case when it meets the bar for one.
New signals like this, once a week. A short email when a case, signal or rule touches software agents. Free; unsubscribe in one click.
Get the weekly emailNewer: Microsoft report: the China-state aligned actor Storm-0940 used the ARTEX agentic framework to automate attacks on hundreds of organizations from September 2026
Older: CVE-2026-114106: HKUDS Vibe-Trading file-read tools