Skip to content
NewA calmer, faster ZeroHour
StoryThreat actorhighExploited in the wild

Cairn AI pentest fork hit at least 19 organizations

3 sources · 3 articlesFirst reported Updated ()1
What's new

November 12: CSO Online adds three-to-seven-stage runs, a shared blackboard, human hints, an authorization.md pretext, and customer-level read, write, and delete access through separate flaws.

Merged summary · x-ai/grok-4.7 · rewritten as coverage arrivesSummaries by Vani AI

Unit 42 says a customized Cairn fork gained access or data at 10 of at least 19 organizations in August–September 2026.

Unit 42 says Chinese-speaking attackers used a customized fork of the open-source Cairn penetration-testing framework, driven by Zhipu GLM and DeepSeek models, for 25 operations against at least 19 organizations in August and September 2026. Access or data theft succeeded at 10 targets spanning semiconductor and advanced manufacturing, defense policy, and civil society. Unit 42 assesses with moderate confidence that the activity is China-nexus collection tracked as CL-STA-1238 and believes one person or a small team conducted it; CSO Online describes the campaign as China-based. Outcomes included more than one million member records at a U.S. teachers’ union, credentials, persistent admin accounts, and an overwritten standards ballot. CSO reports operators could inject hints mid-run and that separate flaws yielded customer-level read, write, and delete access.

Key facts
  • Seven semiconductor, three defense-policy, and four civil-society organizations were among the targets.
  • Thirteen runs used GLM 5.2, eight DeepSeek V4 Flash, and four DeepSeek V4 Pro.
  • A teachers’ union exposure included over one million records and about 31,000 CRM contacts.
  • An open defense-policy endpoint exposed about 500 records, some with .mil and .gov addresses.
  • The fork ran in Hong Kong at 154.44.31.4, with U.S. egress 82.29.129.54.
  • Runs lasted 20 minutes to more than 50 hours, with three to seven stages.
Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat

What each outlet adds

beyond the shared facts
  • Help Net SecuritySector counts, union CRM figure, 500-record endpoint, and password-spraying outcome.
  • Palo Alto Unit 42Model split, run durations, Hong Kong host, and U.S. egress address.
  • CSO OnlineStages, blackboard, human hints, authorization pretext, and separate-flaw access.
Vendors
ZhipuDeepSeek
Threat actors
CL-STA-1238
Malware
Cairn
Organizations
Palo Alto NetworksUnit 42
Countries
ChinaHong KongUnited States
AI models
Zhipu GLM 5.2DeepSeek V4 FlashDeepSeek V4 Pro

Coverage

oldest first · one row per article · UTC
  1. · 9h ago Help Net SecurityFirst report

    Cairn AI pentest framework breached 10 of 19 targeted organizations

    Unit 42 says one operator or a small team hit semiconductor, defense and civil-society targets in two weeks.

  2. · 9h ago Palo Alto Unit 42

    Cairn AI framework used in 25 intrusions against at least 19 organizations

    Unit 42 says access or data theft succeeded at 10 targets, including more than one million records at one organization.

  3. · 9h ago CSO Online

    Unit 42 says attackers used Cairn AI against 19 organizations

    Access succeeded at 10 targets; one business lost more than one million records, Unit 42 reports.