In this edition: a caution from Florida’s courts – and from me – about using AI in the realm of law and legal services. AI is a topic of heated debate, but this update is not about choosing sides between the Luddites and the Techies; this is about being aware, using artificial real intelligence, and getting artificial real, personalized legal advice when you need to protect your business or personal interests.
What Florida’s Courts are Seeing and Saying
As you may know, in order to have any persuasive value, an argument in court must be based on legal authority. If I argue to a court that the law gave my client the right to do what he/she did, I cite to a statute, a case, or other legal authority that says so. This is basic to legal advocacy.
Maybe you’ve already seen the news headlines about attorneys and pro se litigants (people representing themselves in court, without attorneys) embarrassed for using AI and citing to non-existent legal authorities in court filings. But to give you a sense, here are some recent quotes from Florida’s appellate courts about what they are seeing with the use of AI in law.
From Florida’s Sixth District Court of appeal in March:
“Courts across the United States, including Florida’s appellate courts, are currently grappling with an influx of court filings produced by pro se litigants and attorneys alike with the assistance of AI that cite non-existent cases or that cite actual cases for inaccurate legal propositions. “
“Here, the petition filed by Petitioner cites to thirteen cases. Only four of the cited cases both exist and are cited for legal propositions that the cited cases actually represent. Five of the cited cases do not exist. Four of the cited cases are cited for legal propositions that the cited cases do not actually represent. ”
Hessert v. Hessert, — So. 3d –, 51 Fla. L. Weekly D520a (Fla. 6th DCA 2026).
From Florida’s Fourth District Court of Appeal in March:
“Artificial intelligence may assist in research and drafting, but it cannot replace the lawyer’s or litigant’s duty to ensure that every authority cited is real, accurate, and applicable. The responsibility for what is filed rests with the human author, not the machine. ”
Francois v. Vive Financial, LLC, — So. 3d –, 51 Fla. L. Weekly D500a (Fla. 4th DCA 2026).
Did you catch the last part of that quote from the Francois case? “The responsibility for what is filed rests with the human author, not the machine. ” That is the caution from Florida’s courts, and we would all do well to heed it.
What I’m Seeing
Every week I’m seeing decisions like those, where Florida’s appellate courts note this ongoing problem of bad work produced by AI and the burden it imposes on the courts. And courts are sanctioning the offending attorneys and pro se litigants by, among other things, ordering them to pay the opposing party’s attorney’s fees and court costs.
But “hallucinations,” like citations to legal authorities that don’t even exist or that exist but don’t stand for the proposition being argued, are not the only risk of using AI in law.
Another risk is a loss of confidentiality and loss of the attorney-client privilege. Generally speaking, when we use the large language models (e. g. , ChatGPT) – the free versions, at least, and maybe some other versions – we are giving them free training/learning material. (It’s a little bit like feeding a growing teenage boy – he will eat everything you can possibly feed him and still want more!) And that means what we feed them becomes part of the public domain. In other words, we can have no expectation of privacy in the information we feed them.
Why does that matter? When you hire an attorney, you can and should expect your information to be kept confidential. And when you hire an attorney, the attorney-client privilege generally shields the conversations and other communications between you and your attorney and protects them from discovery by your adversaries. But if you go and load details about your legal matter into ChatGPT, for example, you may well be giving up confidentiality. And if you go and load into ChatGPT, for example, information and/or advice your attorney has given you, hoping to get a free second opinion or free follow-up advice, you may well be wiping out the attorney-client privilege and giving your adversaries a basis for arguing that you must hand over communications between you and your attorney that would normally be protected.
Then there are more practical risks. For example, many have said (and I agree based on search results and chats that clients have shown me) that the mainstream AI agents tend to tell you what they “think” you want to hear. That’s bad when it comes to law and legal advice. Legal issues and legal advice are very nuanced, rarely black and white, and call for both knowledge and experience. Part of an attorney’s job is to ask the right questions to understand your situation and be able to give pertinent advice. Another part of an attorney’s job is to tell you like it is, even (and maybe especially) when it’s not what you want to hear.
Sometimes I’m asked to “fix” a document or a situation created by AI or with the use of AI. And sometimes that’s possible. For example, if a client needs a contract prepared, brings me a draft created by or with AI, and asks me to work it up into something appropriate to their situation, I’m often happy to do it. That can be an example of good use of AI in law – as a starter idea or draft for a real lawyer to work up and tailor to a client’s needs. But sometimes it’s not possible; sometimes it’s too late, like when an AI-drafted contract has already been signed and has already led to a dispute.
Conclusion
I’m not here to tell people they shouldn’t use AI. I am here to spread the caution, from Florida’s courts, that “the responsibility for what is filed rests with the human author, not the machine. ” And the caution should extend far beyond court filings to the advice people are seeking and the non-court documents people are preparing. At least in the realm of law, AI needs to be supervised and overruled by an attorney with artificial real intelligence and with artificial real legal training.
