An assistant U.S. attorney in the Eastern District of North Carolina came under fire after a federal judge found that a court filing contained fabricated quotes, misstated case law, and even fake regulatory language.
The person who first spotted the problems was not opposing counsel, but the plaintiff himself — retired Air Force colonel and longtime military lawyer Derence Fivehouse, who is suing the federal government over limits on GLP‑1 coverage under TRICARE for Life.
How the Problems Came to Light
In late 2025, Fivehouse, an experienced military attorney and former staff judge advocate, noticed that a government brief did not read like the sources it cited. The filing, signed by Assistant U.S. Attorney Rudy Renfer, purported to quote several appellate decisions and provisions of the Code of Federal Regulations. When he checked the authorities, some language in quotation marks did not appear anywhere in those opinions or regulations.
Other cites turned narrow, fact‑specific circuit court rulings into sweeping support for the government’s position and described holdings in ways that would surprise anyone who had actually read the cases. To Fivehouse, those were not close calls but misrepresentations. He laid out these discrepancies in his own submissions, forcing the court to confront not just sloppy citations but the possibility that the government’s brief was relying on invented authority.
The Show‑Cause Order
Earlier this month, Magistrate Judge Robert T. Numbers II issued a sharply worded order finding fabricated quotations and misstatements of case holdings in the government’s papers, along with false or misleading statements about how they got there. The court said it had uncovered multiple defective cites to case law and at least two invented quotations attributed to the Code of Federal Regulations, suggesting a pattern rather than a typo.
Renfer told the court he had accidentally filed an unfinalized draft, but after reviewing the record Judge Robert Numbers said he doubted both the accuracy of the brief and the credibility of that explanation. The issue was no longer just a messy document; it was whether a government lawyer could be trusted to explain how his brief was prepared. Because of “the seriousness of these issues,” Judge Numbers ordered senior leaders from the U.S. Attorney’s Office in the E.D.N.C. to appear at a show cause hearing and warned that the office itself could be sanctioned unless it proves this was a one‑off breakdown, not the result of broader supervision failures.
AI Is, Of Course, Behind the Brief
For anyone following AI in litigation, the pattern is familiar: lawyers use public chatbots to draft arguments, paste the output into briefs, and later discover key authorities were AI‑generated. Courts have already sanctioned attorneys for relying on bots that invented cases and citations; now the same concerns are surfacing about filings from within the DOJ.
At the show‑cause hearing in Raleigh, Renfer told Judge Numbers that after accidentally overwriting a prior draft, he “felt panicked” and had AI rewrite the brief, then filed it believing he had reviewed it. Renfer said the decision to use AI was the worst of his career.
Why Fivehouse’s Background Matters
It is not a coincidence that these problems were caught by a pro se plaintiff rather than by opposing counsel. Fivehouse is not a typical self‑represented litigant. He is a retired Air Force colonel, a former staff judge advocate, and a former chief of the legal counsel division in an Air Force base‑conversion agency, with years spent reading dense regulations and formal plans where precise wording matters.
That background helps explain how Derence Fivehouse spotted the issues in the government’s brief. To a casual reader, a plausible‑sounding passage with a string of cites may look authoritative. To someone who has spent decades reading regulatory text and judicial opinions, invented language and mismatched sources can stand out quickly. It also undercuts any attempt to dismiss the errors as harmless or purely technical. From his perspective, this was not a formatting problem; it went to whether the court was being given real law at all.
Fivehouse has called the source of the fabricated text “a material issue,” stressing that the court needs to know whether the words came from the lawyer who signed the brief, from an internal database, or from some undisclosed “helper.”
Consequences for DOJ
U.S. Attorney for the Eastern District W. Ellis Boyle joined Renfer to apologize at the show-cause hearing. The DOJ has not, at least publicly, tied Renfer's exit to any particular platform or product, but the office has since circulated warnings about AI use and referred the matter to the Office of Professional Responsibility. While Judge Numbers could still issue monetary sanctions, suspension from practicing in his court, and even contempt, he did say that he was “heartened” by the steps Boyle is taking.
Judges have long assumed that a federal prosecutor’s citations are particularly reliable. If that assumption no longer holds, they may respond with closer scrutiny and sharper remedies. The Fivehouse episode will likely push DOJ to tighten its AI policies, require real human verification of citations, and treat AI misuse as seriously as any other misquote. It also signals that judges may scrutinize federal filings more closely — and that if language appears in your brief, you own it.
Related Resources:
- New Software for SCOTUS to Detect Financial Conflicts for Justices (FindLaw's Federal Courts)
- SDNY Judge Rules Defendant’s Own Use of AI Tools for Strategizing Not Privileged (FindLaw's Practice of Law)
- New Lawsuit Claims ChatGPT Practices Law Without a License (FindLaw's Practice of Law)
