Part of the AI Regulation News hub.
Florida's Second District Court of Appeal fined an attorney $1,500, made him personally liable for the opposing party's fees and referred him to The Florida Bar for what it called rampant use of hallucinated citations
The appeal itself was affirmed without comment. The court wrote sixteen pages anyway, to explain why an attorney who used a generative AI tool from the beginning of an appeal without any apparent guardrails could no longer plead ignorance, and to set out what it cost him.
Bottom line: Three sanctions, imposed together. A $1,500 fine payable to the clerk within fourteen days. Remand to the trial court to fix the reasonable appellate fees U.S. Bank incurred answering the offending brief, payable by the attorney personally. And a referral to The Florida Bar.
Who this affects: Litigators in Florida's appellate courts, law-firm risk and general counsel setting AI verification policy, professional-responsibility counsel, and any attorney relying on a generative tool for legal research.
Issue date: Decided 21 August 2026. The fine is payable within fourteen days of the opinion. The opinion is marked subject to revision prior to official publication.
What changed: No new rule. The court applied existing professional obligations and made explicit that the ignorance defence is gone, that the attorney may not pass the cost to his clients, and that the fee liability attaches personally.
Analysis: The finding that does the work is evidentiary rather than legal. Because the bulk of the legal argument in the offending filings rested on hallucinated authority, the court concluded it was obvious that generative AI was not the attorney's supplemental research tool, as he claimed, but his primary or only one. That inference - reasoning from the density of fabricated citations back to how the brief was actually produced - is the part other courts can reuse.
Primary sources: Opinion, Fla. 2d DCA, No. 2D2024-1392 (PDF)
- Case
- Capital Standard, LLC; Vladimir O. Tchentsov, as trustee of the TF15529 Land Trust; and Anna Tchentsov, as guardian for Oleg V. Tchentsov v. U.S. Bank National Association, as trustee
- Docket
- No. 2D2024-1392
- Court
- District Court of Appeal of Florida, Second District
- Panel
- Rothstein-Youakim, J., writing; Northcutt and Smith, JJ., concurring
- Below
- Circuit Court for Pinellas County, Michael F. Andrews, Judge
- Jurisdiction
- Florida, United States
- Status
- Affirmed; remanded for determination of appellate attorney's fees. Opinion subject to revision prior to official publication.
- Sanctions imposed
- $1,500 fine to the clerk within fourteen days; personal liability for U.S. Bank's appellate fees on the amended initial brief; referral to The Florida Bar
- Editorial Note
- Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.
- Primary source
- https://flcourts-media.flcourts.gov/content/download/2494161/opinion/Opinion_2024-1392.pdf
What the court actually did
The merits took one sentence. The court concluded that Capital Standard's arguments lacked merit and affirmed the declaratory judgment in favour of U.S. Bank in all respects without further comment.
It then wrote to sanction Capital Standard's counsel, Kenneth Keefe, for his rampant use of hallucinated citations to legal authority in his amended initial brief and subsequent reply brief, and to explain the rationale for the sanctions imposed.
The sequence matters for anyone reading the docket. The client's appeal failed on its own terms. The sanctions run against the lawyer personally, and the opinion is explicit that he is solely responsible for paying the fee award and fine and may not charge his clients for those amounts.
The conduct
The court found that the attorney used a generative AI tool from the beginning of this appeal without any apparent guardrails. The original initial brief was, in the court's words, replete with nonexistent quotations and purported statements of law with wholly inaccurate cites.
What happened next is what elevated the sanction. Shortly after filing, counsel moved to file an amended initial brief on the basis that he had identified critical issues requiring clarification and elaboration. The court observed that ensuring the accuracy of his purported authorities was apparently not among those critical issues, because the amended brief carried the same defect. So did the reply brief.
The court also drew an inference from the scale of the problem: because the bulk of the legal argument in the offending filings was based on hallucinated authority, it was obvious to the court that generative AI was not merely the attorney's supplemental research tool, as he claimed, but his primary or only one.
On his response, the court noted he said he accepted full responsibility for the citation errors and would set up unspecified processes to prevent recurrence, but that he never apologised for wasting the court's and U.S. Bank's time and resources. It added that remorse would not have changed the outcome, citing a Florida decision imposing sanctions even where counsel was deeply remorseful for his lapses of professional judgment.
The ignorance defence is closed
The opinion's framing device is chronological. It begins by noting that it has been three years since Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023), in which a federal court sanctioned attorneys and a firm under Rule 11 for, among other things, citing fake cases created by generative AI and then responding evasively when confronted.
From there the court says it defies belief for a practising attorney to plead ignorance of the risks of relying on generative AI without double-checking the results, and cites the Fifth Circuit in Fletcher v. Experian Information Solutions, Inc., 168 F.4th 231 (5th Cir. 2026): if it were ever an excuse to plead ignorance of the risks of using generative AI to draft a brief without verifying its output, it is certainly no longer so.
It reinforces the point with United States v. McGee, 806 F. Supp. 3d 1264 (S.D. Ala. 2025), that given the broad attention to the problem no attorney can claim ignorance or simply bury their head in the sand when it comes to their own use of AI, regardless of the program used.
The court then adds the sentence that captures its position: and yet here we are.
The Florida line of authority is now long
The most useful part of the opinion for practitioners is its footnote collecting Florida appellate decisions on AI-fabricated authority. It is no longer a short list.
It includes Avery v. Beauzil (Fla. 4th DCA, 27 May 2026), barring a pro se litigant from future filings due in part to AI hallucinations; Gleason v. Marcus, 432 So. 3d 1265 (Fla. 2d DCA 2026), cautioning on the dangers of using generative AI without confirming results; Rodriguez v. Rodriguez (Fla. 6th DCA, 10 April 2026) and Hessert v. Hessert, 431 So. 3d 610 (Fla. 6th DCA 2026), both ordering parties to show cause over hallucinations in certiorari petitions; Gouveia v. Meridian Financial Investments, LLC, 432 So. 3d 49 (Fla. 4th DCA 2026); Russell v. Mells, 426 So. 3d 913 (Fla. 2d DCA 2025), referring an attorney to The Florida Bar; Takefman v. Pickleball Club, LLC, 418 So. 3d 826 (Fla. 3d DCA 2025); Goya v. Hayashida, 418 So. 3d 652 (Fla. 4th DCA 2025), addressing a counterfeit brief; and Gutierrez v. Gutierrez, 399 So. 3d 1185 (Fla. 3d DCA 2024), sanctioning a litigant for fictitious case law.
The court also cites the tracking database maintained by Damien Charlotin, noting that as of the opinion it had identified 497 cases in the United States involving attorneys' reliance on AI hallucinations.
For a Florida practitioner the practical effect of that footnote is that no court in the state now has to write the background section. The authority is assembled, spans all six districts represented in the list, and covers both represented parties and litigants in person.
What the court says it is not deciding
The opinion goes out of its way to isolate the misconduct from the technology. It reiterates that the problem is not the attorney's use of generative AI, quoting the Fifth Circuit's observation that generative AI in law practice may very well be helpful if done properly and carefully, and a Florida federal decision that there is nothing inherently wrong with an attorney properly and competently utilising AI or any of its subsets to practise law or litigate cases.
The violation is defined narrowly: using generative AI without also ensuring that the legal propositions and authority generated are trustworthy.
The court closes the point with a line borrowed from an Indiana federal decision, that the use of artificial intelligence must be accompanied by the application of actual intelligence in its execution.
That distinction is doing real work. An opinion that condemned AI use as such would be weak authority for a bar with firm-wide AI deployments. An opinion that locates the duty in verification is one a professional-responsibility committee can adopt directly.
Three features make this worth circulating internally. The financial exposure is personal and non-billable: the fine and the opposing party's fees fall on the lawyer, who may not charge the client. The evidentiary inference is transferable: a court may reason from the density of fabricated citations that AI was the primary research tool rather than a supplement, regardless of how the lawyer characterises it. And the mitigation that usually helps did not: accepting responsibility and promising unspecified future processes drew an express observation that no apology had been offered for the waste of the court's and the opponent's resources. If your verification policy is a promise rather than a step someone performs and records, this opinion describes what that looks like from the bench.
Source File
https://flcourts-media.flcourts.gov/content/download/2494161/opinion/Opinion_2024-1392.pdf
Open the PDF and confirm four things: the Second District caption and docket No. 2D2024-1392 with the 21 August 2026 date on page 1, the appearance line naming Kenneth H. Keefe of The Keefe Law Firm for the appellants, the second paragraph stating that the court writes to sanction counsel for rampant use of hallucinated citations, and the final page imposing the $1,500 fine, the remand on fees and the referral to The Florida Bar.
We write, however, to sanction Capital Standard's counsel, Kenneth Keefe, for his rampant use of hallucinated citations to legal authority in his amended initial brief and subsequent reply brief and to explain the rationale for the sanctions that we impose. Capital Standard, LLC v. U.S. Bank N.A., No. 2D2024-1392 (Fla. 2d DCA 21 August 2026)
FAQ
What sanctions were imposed?
A $1,500 fine payable to the clerk of the court within fourteen days; a remand to the trial court to determine the reasonable appellate attorney's fees U.S. Bank incurred researching and answering the amended initial brief; and a referral to The Florida Bar. The attorney is solely responsible for the fine and the fee award and may not charge his clients for either.
How was the $1,500 made up?
The court fined the attorney $500 for his belated response to the order to show cause and $1,000 for his AI misuse, imposing $1,500 in total.
Did the client lose the appeal too?
Yes, but separately. The court concluded that Capital Standard's arguments lacked merit and affirmed the declaratory judgment for U.S. Bank in all respects without further comment. The sanctions run against counsel personally.
Did the court criticise the use of AI itself?
No, and it said so expressly. It reiterated that the problem was not the use of generative AI, noting authority that AI may be helpful in law practice if done properly and carefully and that there is nothing inherently wrong with competently using it. The violation was using it without ensuring the legal propositions and authority generated are trustworthy.
Why did the court reject the claim that AI was only a supplemental tool?
Because of the scale. Since the bulk of the legal argument in the offending filings rested on hallucinated authority, the court found it obvious that generative AI was the attorney's primary or only research tool rather than a supplement.
Is there now a body of Florida authority on this?
Yes. The opinion collects decisions from the Second, Third, Fourth and Sixth Districts, including a prior Second District referral to The Florida Bar in Russell v. Mells, and cites a tracking database that had identified 497 US cases involving attorneys' reliance on AI hallucinations as of the opinion.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.