Case gtig-apt45-cve-prompt-farmState actor AI opsWorldwide
North Korea's APT45 fed Gemini thousands of prompts to validate exploits and grow arsenal
GTIG observed DPRK-linked APT45 sending thousands of repetitive prompts to recursively analyse CVEs and validate proof-of-concept exploits, and experimenting with OpenClaw and OneClaw agent frameworks against test labs.
MediumAttack using AIStatus: ContainedVerified case file
Open in the interactive mapCite this case
- Reported
- May 11, 2026
- Location
- Worldwide
- Target
- Systems affected by recently published CVEs
- Severity
- Medium
- Triggered by
- APT45 (North Korea)
- Actor type
- State-sponsored
- Reported loss
- None reported
- CVE
- None
What happened
GTIG observed DPRK-linked APT45 sending thousands of repetitive prompts to recursively analyse CVEs and validate proof-of-concept exploits, and experimenting with OpenClaw and OneClaw agent frameworks against test labs.
Impact: Expanded exploit arsenal; specific intrusions not detailed.
Who triggered it
Actor: APT45 (North Korea)
Attribution: GTIG attributes APT45 to the DPRK with high confidence.
Motive: Exploit development for espionage and revenue operations
How it happened
Attack vector: Bulk automated prompting for CVE analysis and PoC validation
Root cause: Per-prompt safety checks do not stop aggregate exploit research across thousands of benign-looking queries
Timeline
- Thousands of repetitive CVE-analysis prompts observed in Gemini telemetry
- GTIG publishes the case and disables the accounts
Response
Accounts disabled; learnings fed into classifiers.
Current status: Contained
Sources (open in a new tab)
- GTIG AI Threat Tracker May 2026 cloud.google.com
- Secure Blink coverage secureblink.com
Cite this case
North Korea's APT45 fed Gemini thousands of prompts to validate exploits and grow arsenal. AI Threat Watch, case gtig-apt45-cve-prompt-farm, May 11, 2026. https://aithreat.io/cases/gtig-apt45-cve-prompt-farm/
Permanent link: https://aithreat.io/cases/gtig-apt45-cve-prompt-farm/