# Amarsingh v. Frontier Airlines: 10th Circuit sanctions attorney $1,000 for ChatGPT-fabricated citations

> In Amarsingh v. Frontier Airlines, the U.S. Court of Appeals for the Tenth Circuit (opinion by Judge Harris L. Hartz) sanctioned attorney Kusmin L. Amarsingh $1,000 under FRAP 38 after her appellate brief cited seven cases the court could not locate, generated, she admitted, by ChatGPT, and referred her to the Maryland attorney-disciplinary authority.

- Verification status: verified
- Case type: deployment
- Provider: Generative-AI legal drafting (ChatGPT)
- Client: U.S. Court of Appeals for the Tenth Circuit (opinion by Judge Harris L. Hartz): Amarsingh v. Frontier Airlines, Inc., No. 24-1391, Courts / legal (AI-governance honest-negative) (named)
- Sector: legal / US / cross
- Verified on: 2026-08-23
- Canonical URL: https://theinternetninja.com/stories/amarsingh-v-frontier-airlines-10th-cir-hartz-1000-chatgpt-seven-fabricated-citations-frap-38-maryland-referral-2026/
- Source: The Internet Ninja (theinternetninja.com), independent verified-proof platform

## Outcomes

| Metric | Before | After |
| --- | --- | --- |
| $1,000 sanction imposed on the attorney under Fed. R. App. P. 38, payable to Frontier Airlines within 30 days |  |  |
| Seven cited cases were nonexistent; the attorney admitted the citations were generated by ChatGPT |  |  |
| Second sanction: referral to the attorney-disciplinary authority of the State of Maryland |  |  |

## Verification method

Signed federal appellate Order and Judgment (10th Cir.), February 9, 2026, byte-tied to Wayback (CDX digest); quotes verbatim from the order; corroborated firsthand by Colorado Politics and Bloomberg Law.

## FAQ

**How much was the attorney sanctioned in Amarsingh v. Frontier Airlines?**

The Tenth Circuit ordered attorney Kusmin L. Amarsingh to pay $1,000 to Frontier within 30 days under Federal Rule of Appellate Procedure 38, after her opening brief cited seven cases the court could not locate. She admitted using ChatGPT, which generated the seven fabricated citations.

**Did the court refer the attorney for discipline?**

Yes. As a second sanction, the panel directed the Clerk to transmit the order to the appropriate attorney-disciplinary authority of the State of Maryland, where Amarsingh is licensed and listed in active status.

## Full case file

## The problem
Attorneys are increasingly filing court briefs whose case citations were produced by
generative-AI tools and never verified. When those tools "hallucinate," they generate
citations that look real, properly formatted and plausible, but refer to cases that do
not exist. In *Amarsingh v. Frontier Airlines, Inc.*, the Tenth Circuit's review of the
appellant's opening brief "revealed serious flaws, including citations to nonexistent
cases and attribution of propositions or quotations to two real cases that did not stand
for the propositions or contain the quotations" ([source](https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111382504.pdf) [1]).

## What was built
Kusmin L. Amarsingh, an attorney representing herself, used a generative-AI tool to
research and draft her appeal ([source](https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111382504.pdf) [1]).
After the court ordered her to correct the citations or explain them, she admitted she
"had used ChatGPT, which generated the seven fabricated case citations," and that she had
not verified them ([source](https://news.bloomberglaw.com/litigation/appeals-court-sanctions-lawyer-over-ai-hallucinated-errors) [3]; the same admission, "which generated the seven fabricated case citations," appears in the [source](https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111382504.pdf) [1]).
Colorado Politics likewise reported the brief rested on "unverified ChatGPT" citations
([source](https://www.coloradopolitics.com/2026/02/09/10th-circuit-orders-lawyer-to-pay-1000-for-faulty-ai-citations/) [2]).
The panel held that it would "also impose sanctions on Amarsingh for her misuse of
generative artificial intelligence in researching and drafting her appellate brief"
([source](https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111382504.pdf) [1]).

## The outcome
In an unpublished Order and Judgment filed February 9, 2026, the Tenth Circuit affirmed
the dismissal of Amarsingh's claims and, electing "to proceed under Rule 38," concluded
"this appeal is, in part, frivolous as argued" and that her failings were "quite serious
and warrant a significant sanction" ([source](https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111382504.pdf) [1]).
The court ordered that "Amarsingh must pay $1,000 to Frontier to offset attorney fees and
costs Frontier incurred in responding to the problems" in her brief, within 30 days and
with a sworn certification of compliance ([source](https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111382504.pdf) [1]).
As a second sanction, the court directed the Clerk's Office to transmit the order "to the
appropriate attorney-disciplinary authority of the State of Maryland," where she is
licensed ([source](https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111382504.pdf) [1]).
Colorado Politics reported the court ordered the lawyer to "pay $1,000 to the opposing
side" and to "forward the order to the attorney regulators in Maryland" ([source](https://www.coloradopolitics.com/2026/02/09/10th-circuit-orders-lawyer-to-pay-1000-for-faulty-ai-citations/) [2]);
Bloomberg Law reported that "Kusmin L. Amarsingh on Monday was sanctioned $1,000 by the
US Court of Appeals for the Tenth Circuit" ([source](https://news.bloomberglaw.com/litigation/appeals-court-sanctions-lawyer-over-ai-hallucinated-errors) [3]).

## A note on sourcing
The load-bearing evidence here is the strongest kind: a signed, public Tier 1 federal
appellate order that states the $1,000 sanction, the seven fabricated citations, the
ChatGPT admission and the Maryland referral in its own words, not a party's self-report.
The **weakest load-bearing source is Bloomberg Law [3], a subscription/paywalled report
that could not be captured to the Wayback Machine** (its page is login-gated); it is
saved only as a live local capture, so a reader cannot independently re-pull it from a
public archive. That matters little to the record, because every figure it carries is
also stated in the Tier 1 order [1] and in the Wayback-archived Colorado Politics report
[2]; Bloomberg is corroboration, not the sole support for any claim.

## How this was verified
> **Method:** The primary source is the Tenth Circuit's signed Order and Judgment
> (No. 24-1391, filed February 9, 2026, Judge Harris L. Hartz), byte-tied to its Wayback
> capture via a CDX/SHA-1 digest match. Every quoted figure, the $1,000 FRAP 38 sanction,
> the seven fabricated citations, the ChatGPT admission, the Maryland referral, was
> re-checked verbatim against the archived order and against two independent newsroom
> captures (Colorado Politics, Wayback-archived; Bloomberg Law, live capture). No figure
> rests on any party's self-report.
> **Date:** Re-verified 2026-08-18.

## Sources
1. **Tier 1**. U.S. Court of Appeals for the Tenth Circuit · *Amarsingh v. Frontier Airlines, Inc.*, No. 24-1391 · Order and Judgment (unpublished; opinion by Judge Harris L. Hartz) · February 9, 2026 · https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111382504.pdf (archived: https://web.archive.org/web/20260209193841/https://www.ca10.uscourts.gov/sites/ca10/files/opinions/010111382504.pdf)
2. **Tier 2**. Colorado Politics · "10th Circuit orders lawyer to pay $1,000 for faulty AI citations" · February 9, 2026 · https://www.coloradopolitics.com/2026/02/09/10th-circuit-orders-lawyer-to-pay-1000-for-faulty-ai-citations/ (archived: https://web.archive.org/web/20260315194901/https://www.coloradopolitics.com/2026/02/09/10th-circuit-orders-lawyer-to-pay-1000-for-faulty-ai-citations/)
3. **Tier 2**. Bloomberg Law · "Appeals Court Sanctions Lawyer Over AI-Hallucinated Errors" (Sam Skolnik) · February 9, 2026 · https://news.bloomberglaw.com/litigation/appeals-court-sanctions-lawyer-over-ai-hallucinated-errors (subscription/paywalled; no public Wayback capture, held as a live local capture only)

## Related case files
- [Fletcher v. Experian (5th Cir.)](/stories/fletcher-v-experian-5th-cir-hersh-2500-ai-fabricated-citations-frap-46c-inherent-authority-2026/): a sibling federal appellate sanction for AI-fabricated citations, imposed under the appellate rules.
- [Ayinde v. Haringey (UK High Court)](/stories/ayinde-v-haringey-uk-high-court-ewhc-1383-2000-wasted-costs-ai-fabricated-citations-bsb-sra-referral-2025/): the same fact pattern in another jurisdiction, wasted-costs order plus a regulator referral over fabricated authorities.
- [McCarthy v. DEA (3d Cir.)](/stories/mccarthy-v-dea-3d-cir-24-2704-ai-fabricated-citations-public-reprimand-first-impression-2026/): a precedential first-impression appellate response to AI-fabricated citations, choosing a public reprimand rather than a fine.
- [Arabyads v. Alam (ADGM)](/stories/arabyads-v-alam-adgm-cfi-heath-282508-aed-mio-legal-ai-fabricated-authorities-indemnity-costs-2025/): the same conduct penalized commercially, via an indemnity-costs award, in a common-law court outside the U.S.