AI Risk Wire

Cases › made things up › LR-2026-0437

10th Circuit orders lawyer to pay $1,000 for faulty AI citations

Preliminary Why: 1 report: Colorado Politics · subject to change · 1 independent report

Status: no action reported

Severity minor
Reported 9 Feb 2026

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

Severity minor: Test: moderate needs data or systems of others touched with a bounded scope, which the sources do not state; the only effect stated is a $1,000 payment for the opposing party's attorneys' expenses, so the tier is minor.

A self-represented lawyer, Kusmin Linda Amarsingh, used ChatGPT for legal research and drafting in her appeal against Frontier Airlines over a discrimination claim. The filing cited seven cases that did not exist and included a non-existent quote from a case. The Tenth Circuit panel asked her to explain after the airline's lawyers could not locate several of the cited cases. Amarsingh said she had relied on the tool and had not checked the quotations against the case texts. The panel appreciated her honesty and her commitment to being more careful, but ordered her to pay Frontier $1,000 for its attorneys' expenses in responding to the fake citations. The clerk was also directed to forward the order to attorney regulators in Maryland, where she is licensed.

What happened

1 · What went wrongSeven cited cases did not exist, and one quote attributed to a case was not real; the filer said she relied on ChatGPT output.
2 · What it reached or didThe fake citations appeared in a legal filing submitted to the Tenth Circuit.
3 · How it was caughtFrontier's lawyers could not locate several cited cases and flagged it; the panel then asked the filer to explain.

What was done

No source states an action or fix.

Risks

Made-up output relied on
Primary36 incidents

Read from the sources; not yet reviewed by a person.

Controls

Not yet read for controls. Controls that mitigate the primary risk:

Application

Grounding and fact checks · Human approval for high-impact actions · Output filtering and redaction · People told they are dealing with AI

Evidence

No source states: what the AI was doing, when it happened, when it was found, how many records were exposed, a count of the other parties affected, any court case, what set it off, what was done.

Show the full account: 3 items from the primary sources

What happened, in order

  1. August 2024 Magistrate judge recommended dismissing the underlying discrimination lawsuit.
  2. September Frontier's lawyers reported they could not locate several cited cases; the panel ordered Amarsingh to explain.

Not dated in the sources:

  • Amarsingh responded that the seven nonexistent cases resulted from her reliance on generative AI tools.
Show every field, with its source and grade

The AI system

What it was for
legal research and drafting
C one secondary source states it

What happened

Earliest source we hold
2026-02-09
D inferred by us from the described behaviour, reason shown
Who caught it
an outside party
C one secondary source states it
Cause
a human check that existed was skipped or failed
C one secondary source states it
What went wrong
made things up
C one secondary source states it
primary risk
RK14
D inferred by us from the described behaviour, reason shown

Impact

Severity
minor
D inferred by us from the described behaviour, reason shown
Kind of harm
money
C one secondary source states it
Why this severity
Test: moderate needs data or systems of others touched with a bounded scope, which the sources do not state; the only effect stated is a $1,000 payment for the opposing party's attorneys' expenses, so the tier is minor.
D inferred by us from the described behaviour, reason shown

Notes on gaps

The sources report no change made afterwards, only a judicial sanction and a regulator referral, so remediations are empty. The December case involving a different self-represented plaintiff is excluded as a separate incident.

LR-2026-0437 · version 1 · published 2026-11-01 after automatic checks; not yet checked field by field by a person · JSON