Nation-state / AI-driven intrusionHighTLP:CLEAR

Cairn AI intrusions

CL-STA-1238: Autonomous AI Intrusions with the Cairn Framework Against Semiconductor, Defence and Civil Society

Published
12 November 2026
Added
12 November 2026, 18:20 UK
Last updated
12 November 2026

2 to block nowNothing to hunt

Hunt3 platforms

Run the hunt before you block, to find past activity. The same hunt is published for each platform. Preview and copy the queries in the dashboard.

  • Defender XDRKQLDefender portal > Hunting > Advanced hunting. Paste and run over 30 days.Download hunt.kql
  • SplunkSPLSearch & Reporting, last 30 days or more. Reads the CIM data models.Download hunt.spl
  • Google SecOpsYARA-L 2.0Rules Editor. Paste, Retrohunt over 30 days, then turn on live mode.Download hunt.yaral

These queries can return false positives. Know your environment and confirm a hit before acting.

Summary

China-nexus operator running the Cairn AI pentest framework autonomously against semiconductor, defence and civil-society targets. Targets were the semiconductor and advanced-manufacturing supply chain, defence and defence policy, and human-rights and civil-society groups.

Do now

  1. Block the 2 indicators.
  2. Run the hunt.
  3. Block 154.44.31.4 and 82.29.129.54 and search 90 days of firewall, WAF….

Indicators 2

ValueTypeActionCopy
154[.]44[.]31[.]4IPv4block
82[.]29[.]129[.]54IPv4block

Sources Unit 42 · Help Net Security

More detail
Note Single-source indicators (Unit 42); both are attacker-run VPS addresses and may be reassigned later, so review older hits in context. The intrusions used known weaknesses rather than malware, so the IPs are the only published network indicators.

What to do

  1. Block 154[.]44[.]31[.]4 and 82[.]29[.]129[.]54 and search 90 days of firewall, WAF, VPN and identity logs for them, especially from August 2026.
  2. Review internet-facing applications for password spraying and brute force from a single source, and enforce MFA and lockout on partner-integration, CMS and VPN logins.
  3. Look for new administrator accounts, new API tokens and self-registered accounts on internet-facing applications that nobody requested.
  4. Close open self-registration and backend registration endpoints, fix insecure direct object references and mass-assignment flaws, and patch old public CVEs on exposed services.

Affected countries and sectors

Targets
United States, Canada, North America
Sectors
Semiconductors, Manufacturing, Defence, Government / policy, Civil society, Education

Background

Unit 42 reports that from August to September 2026 Chinese-speaking attackers, tracked as CL-STA-1238 and assessed with moderate confidence as China-nexus intelligence collection, ran a customised fork of the open-source Cairn penetration-testing framework with the Zhipu GLM and DeepSeek models to run 25 largely autonomous operations against at least 19 organisations. Targets were the semiconductor and advanced-manufacturing supply chain, defence and defence policy, and human-rights and civil-society groups. The operator gave each task a root domain and a one-line goal, sometimes with VPN gateway addresses, group names and shared passwords gathered beforehand, and the framework planned and ran reconnaissance, web and middleware exploitation, credential collection and persistence on its own for 20 minutes to more than 50 hours. Unit 42 confirmed access or data theft at 10 organisations through weak-password spraying, brute-forced production credentials, unauthenticated account-creation bypasses, insecure direct object references, mass-assignment privilege escalation, exposed cloud storage and old public CVEs, leading to new administrator accounts and a persistent API token in one production system, more than one million member records at another and an overwritten standards-ballot document. The Cairn hosting server and the staging and egress server are on the block list.

Downloads

View folder in repository

MITRE ATT&CK10

Open in the live dashboard Get the blocklist