Elilton v. Meridian Financial: Florida Court Warns Pro Se Litigant Over AI-Hallucinated Citations — and Writes a Limerick About It | Advocate Prakhar

Case at a Glance

Background

Elilton Alves Gouveia and Meridian Financial Investments, LLC were parties to a contract dispute. After litigation, the parties settled, but the settlement itself generated another dispute when a forensic audit of Gouveia’s company revealed findings requiring him to pay Meridian up to $400,000. Meridian moved to enforce the settlement agreement. The Palm Beach County Circuit Court granted Meridian’s motion, and Gouveia — representing himself — appealed.

The appellate court resolved the contract interpretation issue quickly in Meridian’s favour. But the more unusual part of the opinion was what the judges wrote under the heading “AI Spotted.”

The AI Issue

Reviewing Gouveia’s briefs, the court determined that they “were replete with case citations that either do not exist or fail to support the defendant’s arguments.” Specific examples: “Dausch v. Crane, 448 So. 2d 613 (Fla. 4th DCA 1984)” — the case does not exist. “Bennett v. NationsBank, 759 So. 2d 1215 (Fla. 5th DCA 2000)” — the citation leads to a Mississippi case about punitive damages in a tort action, entirely unrelated to the contract issue before the court. Other citations existed but addressed completely unrelated legal questions. The pattern was consistent with a large language model generating plausible-looking but unverified citations.

What the Court Decided

• The trial court’s enforcement of the settlement agreement was affirmed — the contract interpretation was correct [de novo review of settlement agreement].
• The defendant’s briefs showed apparent use of a large language model, producing citations that vanished upon inspection [AI hallucination finding].
• Fake case citations waste the opposing party’s time and money, take the court’s time from other important matters, and can harm the reputation of judges whose names are falsely attached to non-existent opinions [harms of hallucinated authority].
• The court issued a formal warning: future unchecked AI use may result in sanctions under Florida Rule of Appellate Procedure 9.210(c) [prospective deterrence without current penalty].
• Justice Lott concurred separately to call for prophylactic solutions — court-wide AI disclosure requirements — rather than relying solely on post-misconduct sanctions [concurring opinion on systemic response].

“There once was a litigant pro se, / Who let an AI lead the way. / It briefed every claim, / Cited cases—by name, / That vanished by morning’s next day.”

— Limerick generated by ChatGPT 5.2, published in the court’s opinion, Elilton v. Meridian, March 25, 2026

The India Angle

Indian Law Equivalent

Order XLVII of the Code of Civil Procedure, 1908 (Review of judgments) and Order VI Rule 16 (striking out pleadings) provide courts with tools to address frivolous filings. The Contempt of Courts Act, 1971 covers willful submission of false information. For pro se litigants specifically — a growing category in India with the rise of online filing — Section 19 of the Legal Services Authorities Act, 1987 (legal aid) may be relevant: courts might direct AI-dependent litigants to access proper legal aid rather than relying on unverified AI tools.

Bar Council Rules

Bar Council of India Rule 14 (dignity of the court), Rule 15 (fair conduct), and Rule 33 (not to mislead court) apply. Justice Lott’s concurring opinion calling for prophylactic disclosure requirements has a direct parallel in India: several High Courts have begun requiring advocates to disclose AI use in their written submissions, and a uniform BCI rule on this is anticipated.

Practical Advice for Indian Advocates

• When opposing a pro se litigant’s brief, check every citation — the rise of AI means factually invented cases may appear in filings from parties without legal training.
• If your own office uses AI for first-draft research, run every Indian case citation through IndianKanoon or the Supreme Court’s official website before using it — the AI’s Indian law knowledge base is often outdated or incomplete.
• Consider voluntarily including an AI disclosure statement in filings: “No AI tool was used to generate case citations in this submission” or “AI assistance was used and all citations have been independently verified.” This protects your credibility proactively.

Quick Takeaways

• A formal warning today becomes grounds for heavier sanctions on the next offence.
• AI-generated briefs that cite non-existent cases are identifiable and embarrassing — courts will highlight them publicly.
• Prophylactic AI disclosure requirements are coming to courts; stay ahead of them.

Deep Dive: The Limerick Strategy — Why Courts Are Using Humour to Signal Serious Warnings

The Florida Fourth District’s decision to open its AI section with a ChatGPT-generated limerick is not a sign that the court treats the problem lightly. It is the opposite: the humour functions as a sharp contrast to the serious warning that follows. The court is signalling that it sees the hallucination problem clearly, that it is sophisticated enough to recognise AI-generated patterns in briefs, and that future violations will not be treated with the same light touch.

The specific citation failures identified in Gouveia’s brief reveal how LLM hallucinations work in practice. The Dausch v. Crane citation has the right format — a Florida appellate court, a plausible year, a plausible reporter volume — but the case does not exist. The Bennett v. NationsBank citation is more subtle: the case exists, but it is a Mississippi case about tort damages, not a Florida contract case. This is a classic hallucination pattern where the AI correctly identifies a case that exists but assigns it to the wrong jurisdiction, the wrong year, or the wrong legal issue. Both types of error are equally fatal in litigation.

Justice Lott’s concurring opinion is particularly significant for Indian courts to study. Lott argues that the sanction-and-deter model works well for attorneys — who are repeat players, tend to be solvent (so monetary sanctions hurt), and whose bar licenses create a long-term incentive to avoid misconduct. But pro se litigants are not repeat players. Fining them often achieves little because they may lack the means to pay. Lott’s solution: courts should adopt prophylactic disclosure requirements, mandating that any litigant using AI certify that all citations have been independently verified. This shifts the burden upstream — before the hallucinated brief reaches the court — rather than downstream through sanctions proceedings after the damage is done.

For Indian advocates, this distinction matters because Indian courts have a very high proportion of pro se litigants, especially in consumer forums, labour courts, and rent control tribunals. As AI tools become cheaper and more accessible, hallucinated citations in pro se filings will increase. The proactive approach — judicial standing orders requiring AI disclosure and verification certification — is the one that can scale. Indian High Courts, particularly those that have already adopted e-filing systems, are well-positioned to implement this kind of pre-filing certification requirement in their online portals.