Yesterday, I wrote about why artificial intelligence in the legal profession still requires human judgment in AI Makes Good Lawyers Better and Bad Work Faster. My take was that AI can make good lawyers better, but it can also help lawyers produce bad work much faster. I maintained that lawyers should use the machine aggressively, trust it cautiously, and verify it relentlessly.
Today, the Florida Second District Court of Appeal supplied an extraordinary example. 1 The court affirmed a judgment because it found the appellant’s arguments lacked merit. But it wrote sixteen pages for another reason. The court sanctioned the appellant’s lawyer after finding forty-six instances of hallucinated legal authority in an amended initial brief and reply brief.
Thirty-one appeared in the amended initial brief. Another fifteen appeared in the reply. Some involved nonexistent authority. Others were potentially more dangerous because the cases actually existed but did not say what the lawyer claimed they said. Real cases were quoted for language they did not contain or cited for legal propositions they did not support.
That is exactly the problem I wrote about yesterday. A fake case can often be spotted quickly. A real case confidently misrepresented requires somebody to actually read the case. What caught my attention next caused me to do a little research.
My first reaction was that perhaps this involved an inexperienced lawyer who had never developed disciplined research habits. I was wrong. I am intentionally not naming the lawyer here. The published opinion does, and I am attaching it for anybody who wants to read the source. My point is not to add to somebody else’s public embarrassment.
What matters is that this lawyer has been practicing in Florida since 1985. His background includes international trade law, government-related work, international business, real estate, and decades of private practice.
This was not a new lawyer who simply did not understand what lawyers are supposed to do with legal authority. That makes the opinion more troubling and, to me, a little sad.
Artificial intelligence should be enormously valuable to experienced lawyers. Someone who has practiced for forty years possesses decades of context, judgment, pattern recognition, mistakes, victories, and practical knowledge. Put powerful AI in the hands of somebody with that experience, and the combination should be formidable. But experience does not protect us if we stop doing the things experience should have taught us to do.
The facts became worse after opposing counsel pointed out citation problems. Instead of correcting course, the reply brief added another fifteen hallucinations. The court then issued a detailed show-cause order requiring counsel to produce the cited cases, identify where each supported the propositions attributed to it, and explain the failures. More than a month passed without a response.
The lawyer eventually said he had not fully appreciated that generative AI could produce inaccurate or nonexistent citations. The court rejected his description of AI as merely a “supplemental research tool,” concluding that it appeared to have been his primary or only research tool.
The court imposed a $1,500 fine, ordered him personally to pay the opposing party’s attorney’s fees incurred responding to the defective brief, and referred him to The Florida Bar.
The lesson here goes well beyond hallucinated citations. I have been developing a book around the concepts of Contenders and Pretenders. I do not know this lawyer, and I am not going to label somebody a Pretender based upon one terrible episode in a forty-year career. That would be unfair. But I can judge the conduct described in the opinion.
Being a Contender is not a credential we earn once and keep forever. It is reflected in what we actually do.
A Contender does the work. A Contender checks the source. A Contender wants to know whether the case really says what he hopes it says. When something goes badly wrong, a Contender owns it, fixes it, and determines how to prevent it from happening again. This matters enormously in the age of AI because appearances are becoming very easy to manufacture.
A lawyer can have a polished website covering many areas of practice. AI can generate sophisticated practice-area pages, articles, and legal arguments remarkably quickly. The finished product can look authoritative. None of that is mastery.
The lawyer in this case currently advertises insurance and storm claims among a broad range of practice areas. Because insurance law is my field, I looked for publicly available evidence of his insurance experience. I did not find enough to judge its depth or his success in that area fairly.
Absence from an internet search does not establish lack of experience. Most legal work never produces a published appellate opinion. But it caused me to think about a larger problem.
The gap between appearing to possess expertise and actually developing expertise has never been easier to hide.
AI did not create that problem. Lawyers marketed themselves aggressively long before large language models arrived. But AI can dramatically increase the speed, volume, and polish with which the appearance of expertise is produced. This makes real expertise more important, not less. Those selecting attorneys or any professionals should be more mindful of this than ever before.
I use AI extensively and expect to use it even more. It can challenge an argument, locate issues I may have overlooked, organize enormous amounts of information, suggest contrary positions, and force me to reconsider assumptions.
What I cannot outsource is responsibility. If AI gives me a case, I have to read the case. If it gives me a quotation, I have to find the quotation. If it tells me a policy provision means something, I have to read the policy. If it confidently agrees with me, I may be better off asking it to explain why I am wrong or raise the elephant in the room I am not addressing.
The unfortunate thing about these sanction cases is that they risk giving an extraordinary technology a bad name when the technology itself is not the problem. The court expressly made that point. There is nothing inherently wrong with lawyers properly using artificial intelligence. The problem arises when the machine replaces the professional judgment and verification our clients and courts are entitled to expect.
AI should make experienced lawyers better. Forty years of experience should be enormously valuable when combined with artificial intelligence. But only if we continue doing the work that made those forty years valuable in the first place.
Thought For a Friday Afternoon
“The use of artificial intelligence must be accompanied by the application of actual intelligence in its execution.”
— Capital Standard, LLC v. U.S. Bank National Association, Florida Second District Court of Appeal, August 21, 2026, quoting Mid Central Operating Engineers Health & Welfare Fund v. HoosierVac LLC
1 Capital Standard v. U.S. Bank National Assoc., No. 2D2024-1392, — So.3d — (Fla. 2d DCA Aug. 21, 2026).
