Was Thomas v. Commissioner of Internal Revenue sanctioned over AI-generated citations?

Thomas v. Commissioner of Internal Revenue (United States Tax Court, October 23, 2024) is a documented matter involving citations that could not be verified against the reporters named. The recorded outcome was: Pretrial Memorandum stricken. No monetary penalty in this order.

Which AI tool was involved in Thomas v. Commissioner of Internal Revenue?

The order in Thomas v. Commissioner of Internal Revenue does not identify a specific generative tool by name. The court treated the citations as the product of an unverified AI-assisted drafting workflow.

October 23, 2024 · AI sanctions database

Thomas v. Commissioner of Internal Revenue

Thomas v. Commissioner of Internal Revenue is one matter in the Veritas record of US court proceedings involving AI-generated citations that could not be verified against the reporters named. The summary below is drawn from the underlying court document, linked at the foot of this page.

Court: United States Tax Court
Date: October 23, 2024
Party: Lawyer
AI tool: Not identified in the order
Outcome: Pretrial Memorandum stricken
Penalty: No monetary penalty in this order.

What the record shows

The court noticed that several authorities cited in the petitioner's Pretrial Memorandum did not exist, evidencing possible AI hallucinations, and found that two reporter citations corresponded to entirely different cases. Counsel stated a new paralegal had drafted the memorandum, which she did not review; the court deemed the Pretrial Memorandum stricken under Rule 33(b).

This entry reflects the court document as filed. The citations at issue could not be located in the reporters named at the time of the order. The reading here is descriptive of the public record and is not legal advice.

Source documents

Primary court document ↗

The test this filing did not run

Every matter in this record shares one mechanism: a citation that did not resolve to an opinion in the reporter named, in a brief that was filed before any verification step ran. Veritas runs that step. The Filing Risk Scanner extracts each citation, resolves it against reporter and public-record sources, and returns a verdict on each one — located, partial, or not located in reporter — before the brief is signed.

Run a free review Back to the full tracker →

Filed under · AI sanctions database · United States Tax Court

About this publication

Veritas is a pre-filing citation verification platform for law firms and in-house counsel. Every post on the audit log is written by the Veritas editorial team and cross-checked against the underlying sanctions order, standing order, or published opinion it references. Reporter citations are read in their named volumes. Quotations from court orders are checked against the docket.

The product was built by Andrew Zoll, founder of Veritas, who tracks generative AI sanctions in federal and state courts and turns each documented failure mode into verification logic the scanner runs. Veritas was built after the canonical Mata v. Avianca sanctions order made the verification gap a published failure mode of the bar.

AI sanctions database · Sample report · Pricing · Run a scan