Observed incident
BUGSTER used AI assistants in university credential theft and exfiltration
Reported
Published here
What happened
Unit 42 traced BUGSTER’s use of Hermes Agent, DeepSeek V4 and Cursor to university attacks in India, including a reported database theft exceeding 5,200 personal records.
Parties and reporting
AI tools and systems: Hermes Agent, DeepSeek V4, Cursor IDE
Outcome
Unit 42 reported confirmed exfiltration from one university and a candidate login from another. VoIP fraud remained planning or early setup; containment and victim recovery were not disclosed.
Operating lesson
Require strong administrator authentication, detect password guessing, restrict database access and monitor bulk data exports.
Evidence and limits
Findings derive from the actor’s operational environment. This is distinct from Unit 42’s earlier Chinese-speaking actor. Model configuration alone does not prove every configured model participated; cryptocurrency follow-on attacks were not observed.
Sources
- Unit 42 investigation of BUGSTER’s AI-assisted activities
Primary source ·