sa国际传媒

The International Rescue Committee Inc. asked a Texas federal court to sanction a former worker and counsel for "poison[ing] the evidentiary well" by using ChatGPT to tamper with documents produced for discovery, according to a brief and motion for sanctions.

Rawa Hiwedi, who sued the International Rescue Committee for overtime pay under the Fair Labor Standards Act, admitted to creating and manipulating documents using ChatGPT, an artificial intelligence large language model, including two grocery store receipts Hiwedi claimed showed overtime work with client families, according to the brief and motion filed Wednesday.

"Defendant's only recourse is to proceed as if all of plaintiff's evidence is potentially fabricated," the brief said. "Plaintiff's misconduct has irrevocably contaminated the evidence and massively prejudiced defendant. Sanctions up to and including dismissal with prejudice are warranted."

The International Rescue Committee said that Hiwedi and her counsel violated the Federal Rules of Procedure as well as local Texas rules and legal ethics.

There are several contributing factors to the severity of Hiwedi's violations, the committee said. Hiwedi knowingly used ChatGPT, but initially denied doing so. Only after repeated questioning did she relent.

Under the Federal Rules of Procedure, attorneys must also review the plaintiff's document production and such a review by Hiwedi's counsel "would have caught the clearly distorted receipts that had all the hallmarks of AI-gibberish," the brief said.

"Texas Lawyer's Creed is also relevant," the brief said. "Texas-barred lawyers agree to be conscious of their duty to the judicial system, conduct themselves in a professional manner, and refrain from knowingly misrepresenting, mischaracterizing, misquoting, or misciting facts or authorities to gain an advantage."

Two of the receipts that were produced for discovery had clear AI hallucinations, the committee said, including "non-sensical" words like "TRANSACTION RETTESENEO," "CASH DACK ANGUNT" and "GROGERY."

Another factor supporting dismissal and sanctions, the committee said, is the fact that the plaintiff admitted to not knowing which or how many documents were manipulated by AI.

The committee, therefore, could not trust the veracity of future discovery documents, and Hiwedi's action prejudice the committee's ability to mount a defense, the brief said.

"Plaintiff's admission that she used ChatGPT to modify other documents, combined with her inability to recall what documents were modified, clouds the provenance every single document she produced," the brief said. "And Plaintiff's failure to disclose this conduct until after being repeatedly pressed on it in her deposition renders any future productions suspect."

Counsel for the plaintiff and the committee did not immediately respond to requests for comment on Friday.

Hiwedi is represented by Vincent J. Bhatti and Ditty S. Bhatti of The Bhatti Law Firm PLLC.

The International Rescue Committee is represented by Jonathan E. Clark and Graydon A. Cowan of Sheppard Mullin Richter & Hampton LLP.

The case is Hiwedi v. International Rescue Committee Inc, case number 3:25-cv-00918, in the U.S. District Court for the Northern District of Texas.

--Editing by Alex Hubbard.

For a reprint of this article, please contact reprints@law360.com.