Over $24K in Sanctions for Alleged AI Hallucinations in Case Against FIFA

For more on the underlying errors in the plaintiffs' brief, see here. An excerpt from yesterday's sanctions order, in Puerto Rico Soccer League NFP, Corp. v. Federacion Puertorriquena de Futbol:

Plaintiffs' motions included at least fifty-five defective citations, requiring hours of work on the Court's end to check the accuracy of each citation. Plaintiffs' counsel never offered a satisfactory explanation for why their citations in multiple motions were so severely flawed. Plaintiffs denied using generative artificial intelligence. But the sheer number of inaccurate or nonexistent citations suggests otherwise. And in any event, the violations of Fed. R. Civ. P. 11 and applicable ethical rules occurred regardless of whether they were caused by misuse of generative artificial intelligence or other means. This behavior stands in contrast to several of the cases [that imposed much lower sanctions], where various attorneys facing sanctions offered an explanation as to how they erred.

Defendants actually asked for $60K in compensation for the attorney fees and research costs expended to deal with the incorrect citations, but the court reduced that by 60%:

The Court is aware that the errors committed by Plaintiffs' counsel received national attention, and that given both attorneys work for small firms and describe themselves as solo practitioners, the initial lodestar amount would prove a heavy financial burden. Furthermore, it is well-known that "an appropriate sanction should be no more severe than necessary to assure the deterrent objective" of Fed. R. Civ. P. 11…. [T]he Court is aware of no cases where a sanction approaching sixty thousand dollars for the misuse of artificial intelligence has been applied.

Salvador J. Antonetti-Stutts and Aníbal A. Román Medina (O'Neill & Borges LLC) and John J. Kuster, Jon Muenz, and Amanda M. Blau (Sidley & Austin LLP) represented defendants on the sanctions motion.

Comments

[MickeyFitz]

Will they ever learn?

[David Nieporent]

Not sure which failure-to-learn you're referring to: the use of the AI, or the denial of the use of AI after they were caught:

Plaintiffs denied using generative artificial intelligence. But the sheer number of inaccurate or nonexistent citations suggests otherwise. And in any event, the violations of Fed. R. Civ. P. 11 and applicable ethical rules occurred regardless of whether they were caused by misuse of generative artificial intelligence or other means.

Why deny it? The alternative is that someone consciously fabricated the various cases. Which would be even worse.

[Stupid Government Tricks]

given both attorneys work for small firms and describe themselves as solo practitioners, the initial lodestar amount would prove a heavy financial burden.

My IANAL question is, if the small firm finding the hallucinations actually did spend that many billable hours in their investigation, why should they have to eat any of the cost? Why shouldn't all that cost be owed by the hallucinating firm?

Or did the investigating firm pad their bill, or intentionally ask for more as a deterrent? Since the hallucinating firm seems to be denying any responsibility, isn't a deterrent factor reasonable?

[Reallynotbob]

It's often an hourly rates adjustment. Looks like the innocent party had expensive lawyers. It's fine for a party to choose to spend on a Cadillac firm but Courts usually cut the total to Chevy rates when the party being sanctioned isn't a big firm/corporation. (I have not investigated the relationship of the parties or nature of counsel beyond what was in the original post, so this is just a general response.)

[pdshaw]

The phrase, "both attorneys work for small firms . . .," refers to two plaintiffs attorneys representing multiple parties.

On the other side of the ledger, there were fifteen defense attorneys (and one paralegal) from five different law firms, who were also representing multiple parties, who billed on the hallucination issue.

I think the court here recognized there was a collective action issue in which one or two lawyers could have checked the citations and filed a motion at a cost of no more than $12k, particularly given it was their job to check opposing counsel's citations to begin with. But because the law firms acted independently, the costs are higher.

The amount of the award is not intended to make the movant whole, but present a deterrent to future misconduct. I would not be surprised if the framing of the ruling is intended to offer a deterrent against multiple law firms in the future running up legal bills with the expectation that their client will be reimbursed. Not saying that happened here, but the thought had to have crossed the court's mind (or their clerks) while reviewing all of the statements of legal costs that "wouldn't this be simpler if only . . .."

[JonFrum]

"[T]he Court is aware of no cases where a sanction approaching sixty thousand dollars for the misuse of artificial intelligence has been applied."

Or: It's never been done before, so it can't be done. So in the first AI misuse case, how did the court apply ANY sanction? It had never been done before. Muddled thinking.