THE PERILS OF AI CATCH A LAZY FLORIDA LAWYER | EDITORIAL

Florida teachers alerting their students to the pitfalls of relying on artificial intelligence could do worse than assign them to write a paper about an unfortunate St. Petersburg lawyer named Kenneth Keefe.

He’s in a heap of trouble for his “rampant use of hallucinated citations” in a brief he filed with the Florida’s Second District Court of Appeal.

He not only lost the case, but a three-judge panel fined him $1,500, left him on the hook to pay the other side’s lawyers for what it cost them to research and refute his brief — and referred him to the Florida Bar for possible discipline. Keefe’s case is stirring lawyer-to-lawyer chatter on social media.

“Attorney Keefe used a generative AI tool from the beginning of this appeal without any apparent guardrails,” said the opinion by Florida Judge Susan Rothstein-Youakim. His “original initial brief was replete with nonexistent quotations and purported statements of law with wholly inaccurate cites.”

Keefe then filed an amended brief to deal with what he called “’critical issues that require clarification and elaboration,’’’ the judge wrote. “Apparently, however, ensuring the accuracy of his purported authorities was not among those ‘critical issues’” because the amended brief “included at least 31 instances of Attorney Keefe’s citing or quoting existing cases for propositions they do not contain.”

Citing a non-existent case

Lawyers for U.S. Bank, whose trial court victory Keefe was appealing, filed what the judge called a “remarkably tempered response.” Keefe’s reply added “another 15 AI hallucinations, including the fake case of Board of Regents v. Wilson, 365 So 2d 213.”

Keefe’s second brief cited “several of the offending cases from the amended initial brief for additional propositions that are nowhere to be found in them; and, for good measure, citations to several new cases with that same problem,” the court said. The Second District said it was difficult to contact Keefe to have him explain.

“Rather than immediately fall on his sword, he ignored us for more than a month,” the court said, responding only to a new order directing him to tell his clients what was going on.

His eventual reply: “In preparing portions of those filings, I utilized artificial intelligence as a supplemental research tool. I did not fully appreciate that such tools can generate inaccurate or nonexistent citations unless each authority is independently verified through reliable legal research sources.”

The court said it was obvious that generative AI “was not merely Attorney Keefe’s ‘supplemental research tool,’ but his primary or only one.”

Should have known better

The court emphasized that he — and all lawyers — should have known better. Three years ago, a federal court in New York punished lawyers and a law firm for citing fake cases supplied by generative artificial intelligence. That news was reported widely, and similar cases have followed, including last year in Florida, when lawyers for the massive personal injury law firm Morgan & Morgan were fined thousands of dollars by a federal court for their AI-generated hallucinatory briefs.

“We may not be so gentle on attorney offenders in the future,” the court warned.

The problem, the Florida appellate court said, was not that Keefe used AI for research, which it acknowledged can be useful, but that he had not used what another court called “actual intelligence” in its application.

That should be a cautionary tale not just for lawyers, who have long used search engines — as newspapers do — for research and relevant examples. Anyone in any field must be responsible for the accuracy of their sources, and that goes double where life and safety are involved. A mistake by an AI supercomputer would be no less lethal for prescribing the wrong medication than if it were, say, a doctor’s human error.

In this Florida case, opposing lawyers were responsible for fact-checking Keefe’s brief. But no such adversarial system exists to catch errors students may extract from AI in lieu of doing their own research and writing. Their instructors can’t be AI detectives, and AI detection tools are often inaccurate.

The moral hazard in academic cheating is the same as it was when students paid to have real people ghost-write their term papers, and it’s just as wrong and ultimately self-destructive. They only learn how to cheat.

The New York Times reported on AI’s impact on the livelihoods of thousands of college grads in Kenya, a country of high unemployment. They were writing papers for college students in the U.S. and around the world. Many are now out of work and others are earning fees for rewriting papers generated by AI to make the authorship appear human.

To any teacher who might assign students to write about this, here is a suggestion: Require them to bring their notes to class and compose the essays there — no laptops allowed. As technology improves at warp speed, it’s getting too easy to cheat.

The Sun Sentinel Editorial Board consists of Opinion Editor Steve Bousquet, Deputy Opinion Editor Dan Sweeney, editorial writers Pat Beall and Martin Dyckman, and Executive Editor Gretchen Day-Bryant. To contact us, email at letters@sun-sentinel.com.

Legal Disclaimer:

EIN Presswire provides this news content "as is" without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.