Joe Exotic's Appeal Gets the AI Hallucination Treatment, and the Tenth Circuit Noticed

Joe Exotic’s lawyer filed an appeal brief in the Tenth Circuit that contained a citation the court could not find anywhere. The panel’s response, as reported by Above the Law this week, was a sentence that should be tattooed on the inside of every practitioner’s skull: “To our eye, the error has all the hallmarks of a so-called AI ‘hallucination,’ a circumstance where an AI large language model generates an output that is fictional, inaccurate, or nonsensical.”

That is the Tenth Circuit politely telling counsel that the court thinks ChatGPT wrote part of the brief.

The case is United States v. Maldonado-Passage, the ongoing saga of Joseph Maldonado-Passage (Joe Exotic) and his effort to unwind a 21-year federal sentence for a murder-for-hire plot against Carole Baskin. He has been litigating in the Tenth Circuit since 2021, with prior appeals and sentencing remands bouncing back and forth. This latest filing is part of that continuing fight. His legal team is now the latest in a growing list of lawyers publicly flagged by a federal court for filing something that looks synthetic.

Why this one matters more than the last ten hallucination stories

We are past the novelty phase. Courts have been sanctioning lawyers for AI-generated fake citations since the Avianca fiasco in 2023. We have seen six-figure sanctions in Oregon. We have seen Sullivan & Cromwell apologize in the New York Times on April 21 for submitting a brief with fake citations generated by AI. A BigLaw firm with the deepest resources in the profession still had it happen.

What makes Joe Exotic’s situation different is the spotlight. This is not an obscure Rule 12 motion in a civil dispute nobody is watching. It is a high-profile federal criminal appeal involving one of the most recognizable inmates in America. The Tenth Circuit’s order will get reprinted, summarized, and circulated in CLEs for the next year. Any lawyer who reads legal news now has a fresh data point: even in the cases that attract national press, counsel is still letting AI write the citations.

What the court’s phrasing tells you about where this is heading

Note the careful construction. The panel did not flatly accuse counsel of using AI. It said the error has “all the hallmarks” of a hallucination. That is the court leaving counsel room to respond, while making clear that the inference is obvious.

This is the direction every federal court is moving. They will not need to prove the lawyer used AI. The fact pattern, a citation that does not exist, a fabricated quote, a phantom judge, will be enough to trigger an order to show cause. Counsel will then be in the position of either admitting AI use or explaining some other mechanism by which a fake citation appeared in a federal filing. Neither answer is good.

The Second Circuit, the Ninth Circuit, and district courts in Texas and Illinois have all now issued standing orders or individual rulings requiring disclosure of AI use in briefs. The Tenth Circuit’s language in the Maldonado-Passage order is the soft version. The hard version is coming.

The “I didn’t know my paralegal used it” defense is dead

Every sanctioned lawyer in these cases tries some version of the same story. I delegated the research. I trusted the associate. I did not personally run ChatGPT. The courts have stopped accepting it. Federal Rule of Civil Procedure 11 and its criminal counterparts impose a non-delegable duty to verify. The signature on the brief is the signature on the brief. There is no “my tool did it” safe harbor.

If you are a partner, a solo, or a GC outsourcing to outside counsel, the practical consequence is that you own the output of whatever AI your firm is using. If the associate runs the research through an AI that hallucinates, the partner who signs the brief is still the one who answers the show-cause order. The reputational damage lands on the named attorney, not the vendor.

What the firms doing this right are actually doing

The firms that have not had a public AI hallucination incident are not the firms that banned AI. Banning AI just means your associates use it at home and do not tell you. The firms avoiding sanctions have three controls in place:

1. A mandatory second-source verification rule. Every citation in every filing, regardless of how the research was conducted, is pulled in Westlaw or Lexis and verified before filing. No citation goes in a brief that has not been opened and read in a primary-source database.
2. A designated AI-permitted workflow. Associates use approved tools (Harvey, CoCounsel, Lexis+ AI) only, not consumer ChatGPT. The approved tools are integrated with citation verification and do not generate free-form case law from a prompt.
3. Written disclosure protocol. When AI is used in drafting, counsel files an acknowledgment to that effect, which most federal judges now either require or strongly prefer. The disclosure is cheap. The sanctions are not.

What to do now

If you are in a litigation practice:

• Pull every brief your firm has filed in the last 60 days. Spot-check ten citations per brief against Westlaw or Lexis. If any are fake, you have a problem you want to find before opposing counsel does.
• Put a written verification policy in place this week. Not a memo. A checklist attached to every filing signoff.
• Assume opposing counsel is now running your briefs through their own citation-checking tools. Several vendors launched AI-hallucination detectors in Q1 2026. The risk of being caught has gone from “maybe” to “almost certain.”

If you are a GC hiring outside counsel:

• Ask every firm on your panel to show you their written AI use policy. If they do not have one, that is the answer.
• Include an indemnification clause in engagement letters for sanctions arising from AI-generated errors. The firms that are confident in their process will sign it without a problem.

The Joe Exotic filing will not be the last one to end up in a published opinion. It is the latest reminder that the enforcement curve is steepening, not flattening. Courts that spent 2024 and 2025 issuing warnings are now using phrases like “all the hallmarks.” The next phase is fines, suspensions, and referrals to state bar disciplinary counsel. We are already seeing all three in 2026.

Verify your citations. Sign your own briefs. Do not let a language model put your bar card on the line.