CL-STA-1238 Cairn autonomous intrusions
CL-STA-1238 ran 25 autonomous AI intrusions with the Cairn framework against semiconductor and defense targets
Posted · Spectrum Delta Research Team
Summary
A China-nexus operator tracked as CL-STA-1238 used a customized Cairn AI penetration-testing framework running Zhipu GLM 5.2 and DeepSeek V4 models to autonomously conduct 25 intrusion operations against at least 19 semiconductor, defense and civil-society organizations in August and September 2026, gaining unauthorized access or data retrieval at 10 of them, Unit 42 reports (B/2). Operators supplied only a root domain and a one-line goal; Cairn agents planned and executed three-to-seven-stage operations lasting 20 minutes to over 50 hours, benchmarking models and escalating to DeepSeek V4 Pro against targets that resisted DeepSeek V4 Flash. Confirmed gains include more than one million member records and staff password hashes at a US teachers' union, roughly 500 user records including .mil and .gov accounts at a defense-policy identity platform, unauthorized administrator accounts and a persistent API token at a semiconductor company, and an overwritten live standards ballot. Unit 42 assesses with moderate confidence that the activity is China-linked intelligence collection, likely run by a single operator or small team.
Breakdown
Countries
- United States
- Canada
Targets
- Industry: Semiconductor manufacturing
- Industry: Defense and defense policy
- Industry: Human rights and civil society
Technical exposure
- Internet-facing applications accepting weak, default or shared group passwords
- Unauthenticated account-creation endpoints in content-management systems
- Open backend registration endpoints on identity platforms
- Unauthenticated write access to cloud storage holding business documents
- Credentials valid across multiple accounts or partner integrations
Adversary
CL-STA-1238 (China-nexus, Unit 42 moderate confidence) · Unit 42 (B/2)
Procedures (4)
Cairn agents autonomously map the attack surface
- T1595.002Active Scanning: Vulnerability Scanning
Logs
- Web server access logs · Sustained automated requests enumerating paths, parameters and subdomains, sourced from 82.29.129[.]54 (US staging and attack egress) or 154.44.31[.]4 (Hong Kong Cairn hosting)
WAF event logs · Scanning and probing patterns against exposed applications from the Cairn staging or hosting IPs
Detectors
- Advanced IP Defense · Blocks traffic from known malicious source IPs, including attacker staging infrastructure
Password spraying yields dozens of valid credentials
Logs
- Web application authentication logs · Many failed logins across accounts from one source followed by successful logins; a single production credential valid on multiple accounts (semiconductor partner-integration application)
- Application audit logs · Dozens of valid credentials harvested from an internet-facing application server; subsequent access to non-public internal-directory databases
Detectors
- Cortex XSIAM behavioral analytics · Embedded AI engine analyzes user, system and security events to generate prevention rules for anomalous activity
Unauthenticated account creation and rogue admin accounts
Logs
- CMS audit logs · Account or session creation without a completed registration flow (teachers' union CMS account-creation bypass); bulk member-data reads under the new sessions
- Application audit logs · New administrator accounts and API tokens created outside change windows (semiconductor production system); open backend registration endpoint exposing user records
Bulk data retrieval and ballot document overwrite
Logs
- Database and application audit logs · Bulk reads: more than one million member records and about 31,000 CRM contacts (teachers' union); roughly 500 user records including .mil and .gov accounts (defense-policy identity platform)
- Cloud storage audit logs · Unauthenticated write overwriting a live standards-ballot document (object-storage data events)
Indicators of compromise
IPs (2)
- 154.44.31.4Cairn hosting server, Hong Kong, AS979 NetLab Global (Unit 42); hunt-only
- 82.29.129.54staging server and attack egress, United States, AS142036 Hosteons (Unit 42); hunt-only
Sources
- B/2Unit 42: Attackers Use Autonomous AI to Target Semiconductor, Defense and Civil Society Sectors
- C/3Check Point Research: 16th November Threat Intelligence Report (retelling)
Grades are source reliability (A–F) and information credibility (1–6).
Questions, detections or sightings of CL-STA-1238 Cairn autonomous intrusions? Bring them to the community.