Incident·WS-0062·Nov 5, 2025·in the wild·AI-powered cyberattacks

Malware starts calling AI models mid-attack

Google Threat Intelligence Group / PROMPTFLUX (Gemini API), PROMPTSTEAL (APT28, Qwen2.5-Coder)

53/100
WS-3 Serious
Harm score 45, signal score 62. Autonomy: acted (the AI took a bounded action). How the scores work.
Harm 2/5
damage actually done
Reach 2/5
who was affected
Autonomy 2/4
how much the AI did on its own
Novelty 4/5
how new the behavior was
Permanence 3/5
how hard it is to undo

GTIG reported malware families that call LLMs at runtime: PROMPTFLUX, an experimental VBScript dropper asking Gemini to rewrite its own code for evasion, and PROMPTSTEAL, used by Russia's APT28 against Ukraine to have Qwen generate data-theft commands. Google disabled associated assets and noted the capability is still nascent.

Damage

PROMPTFLUX: none realized (in development); PROMPTSTEAL: data theft in Ukraine operations, unquantified

Sources

Google, second source. PROMPTSTEAL was reported earlier by CERT-UA as LAMEHUG (July 2025). Some researchers called PROMPTFLUX overhyped since self-modification code was commented out.

Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.