Cases › security breach › LR-2026-0744
Russia-aligned UAC-0099 prompt injection against AI malware analysis (GuardBreaker)
Confirmed Why: 2 independent outlets · 2 independent reports
Status: no action reported
No source states: loss or records, developer and model, agent tool, could do, deployer, industry.
Severity minor: Test: contained to one deployer or evaluation, no third party touched. The source describes an intrusion against a single target in Ukraine.
A Russia-aligned threat actor tracked as UAC-0099 embedded a text comment in a malicious VBScript aimed at an AI-assisted malware analysis workflow. The comment, which asked for help building a nuclear weapon, was meant to trigger a language model's safety mechanisms so that it would stop analysing the rest of the code. The activity was seen during an early-stage intrusion against a target in Ukraine. ESET, a security company, identified the technique, which it named GuardBreaker, and reported it publicly. The researchers said that an AI scanner refusing or abandoning a file could leave a blind spot, and recommended that AI results be cross-checked with other tools and human review.
What happened
No source tells the story step by step yet.
What was done
No source states an action or fix.
Risks
Primary50 incidents
Rising: 6 vs 2.8 a week
Also96 incidents
Rising: 24 vs 4.0 a week
Sorted by rule from the case's fields; not yet read from the sources.
Controls
Not yet read for controls. Controls that mitigate the primary risk:
HarnessOutside content treated as data, not instructions · Prompt-injection detection · Confirmation after untrusted input · Validation of tool inputs and outputs · Tool allow-list
ApplicationHuman approval for high-impact actions · Least-privilege access for the agent · Segmentation of agent components · Alerts on unusual tool use
Evidence
- What set it off hidden instructions in its input
No source states: a step-by-step account, when it happened, when it was found, how many records were exposed, the direct loss, a count of the other parties affected, any court case, the cause, what was done.
Show every field, with its source and grade
What happened
Impact
Context
Notes on gaps
The sources do not say whether the injected text succeeded in stopping the analysis. No remediation by the affected party is reported.
LR-2026-0744 · version 1 · published 2026-11-01 after automatic checks; not yet checked field by field by a person · JSON