Cases › copyright or IP violation › LR-2026-0396

Attempted model extraction of Google Gemini through more than 100,000 prompts

Preliminary Why: 1 report: Ars Technica · subject to change · 1 independent report

Status: fix reported

Severity minor

Reported 12 Feb 2026

Developer and model Google · Gemini

No source states: loss or records, agent tool, could do, deployer, industry.

Google's Gemini AI chatbot was the target of attempts to clone its knowledge by prompting it, with the activity attributed to commercially motivated actors. One adversarial session reportedly prompted the model more than 100,000 times across various non-English languages, collecting responses ostensibly to train a cheaper copycat. The chatbot responds to anyone online, which the attackers exploited through repeated queries. Google's internal security teams flagged the activity, and it was linked to actors seeking rival models by copying core behaviour patterns. Once identified, offending accounts were blocked and new safeguards were implemented to limit how much sensitive reasoning information can be inferred through repeated queries. Google describes the activity as model extraction and considers it intellectual property theft, a position the source calls somewhat loaded.

What happened

1 · What it was doingAttackers repeatedly questioned Gemini to map its internal logic and reconstruct how it responds.

2 · How it was caughtGoogle's Threat Intelligence Group flagged the activity.

What was done

Offending accounts were blocked and new safeguards were added.

Fix reported

“Offending accounts blocked and new safeguards implemented to limit sensitive reasoning information inferred through repeated queries”

Same developer, same risk, after the fix: LR-2026-0222

Risks

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:

Model

Application

Output filtering and redaction

Similar incidents

ICMP evidence that Google Gemini reproduced lyrics by Australian artists2026-10-19

Gemini lock-screen flaw lets strangers send messages from locked Android phones2026-07-17

Google Gemini global outage exceeds seven hours2026-06-11

Previously harmless Google API keys expose Gemini AI data2026-02-26

All cases of this kind ›

Evidence

No source states: what went wrong, what it reached or did, 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 set it off.

Show the full account: 2 items from the primary sources

What happened, in order

Not dated in the sources:

Google Threat Intelligence Group flagged the extraction activity.
Offending accounts were blocked and new safeguards were implemented.

Show every field, with its source and grade

The AI system

Model

Gemini

C one secondary source states it

Developer

Google

C one secondary source states it

What happened

Earliest source we hold

2026-02-12

D inferred by us from the described behaviour, reason shown

What was done

Offending accounts blocked and new safeguards implemented to limit sensitive reasoning information inferred through repeated queries

C one secondary source states it

What went wrong

D inferred by us from the described behaviour, reason shown

Impact

Severity

minor

D inferred by us from the described behaviour, reason shown

Notes on gaps

The specific safeguards added are not described, so no preventive control is credited.

LR-2026-0396 · version 1 · published 2026-11-01 after automatic checks; not yet checked field by field by a person · JSON