Case gtig-gemini-state-misuseState actor AI opsWorldwide
Google: state hackers use Gemini across all attack stages; 100K-prompt extraction try
GTIG reported Chinese, Iranian, North Korean and Russian groups using Gemini for recon, exploit research and lures, and blocked a 100,000+ prompt campaign to extract its reasoning.
MediumAttack using AIStatus: ContainedVerified case file
Open in the interactive mapCite this case
- Reported
- Feb 12, 2026
- Location
- Worldwide
- Target
- Google Gemini / multiple sectors
- Severity
- Medium
- Triggered by
- APT31, Temp.HEX, APT42, UNC2970 and Russian groups; unnamed distillers
- Actor type
- State-sponsored
- Reported loss
- None reported
- CVE
- None
What happened
GTIG reported Chinese, Iranian, North Korean and Russian groups using Gemini for recon, exploit research and lures, and blocked a 100,000+ prompt campaign to extract its reasoning.
Impact: No breakthrough capabilities observed; Google says extraction attempt was flagged in real time and reasoning traces protected. Accounts and infrastructure disabled.
Who triggered it
Actor: APT31, Temp.HEX, APT42, UNC2970 and Russian groups; unnamed distillers
Attribution: Attributed by Google Threat Intelligence Group; distillation campaign actors not publicly named
Motive: Espionage support (recon, lures, exploit research) and intellectual property theft via model extraction
How it happened
Attack vector: Abuse of legitimate Gemini accounts and APIs; reasoning trace coercion prompts to extract internal chain of thought
Root cause: General-purpose AI assistants accessible to state actors; API access enables large-scale output harvesting for distillation
Timeline
- Previous GTIG AI Threat Tracker covers PROMPTFLUX-era findings
- GTIG publishes report on state misuse and a 100,000-plus prompt extraction campaign
Response
Google disabled abusive accounts and infrastructure, hardened Gemini classifiers and warned of takedowns and legal action.
Current status: Contained
Sources (open in a new tab)
- BleepingComputer bleepingcomputer.com
- SecurityBrief securitybrief.co.uk
- TechInformed techinformed.com
Cite this case
Google: state hackers use Gemini across all attack stages; 100K-prompt extraction try. AI Threat Watch, case gtig-gemini-state-misuse, Feb 12, 2026. https://aithreat.io/cases/gtig-gemini-state-misuse/
Permanent link: https://aithreat.io/cases/gtig-gemini-state-misuse/