Jack Russo of Computerlaw Group LLP's arguments came during a hearing held via Zoom before U.S. District Judge Jeffrey S. White on an order to show cause for why the court should not grant MPH International's administrative motion to withdraw its motion for summary judgment, which contained roughly a dozen AI-generated citation errors.
The filing came in an otherwise small-time copyright and contract dispute between mobile app platform Support Community Inc. and the Houston, Texas-based software maker MPH International over proprietary software that allows a user to service its customers with certain types of online communications.
Before the hearing, Judge White issued a tentative ruling that said he is inclined to grant the motion to withdraw in part so that defense counsel can replace the problematic motion with a corrected version, and so that neither side has to spend additional time working on the briefs.
However, Judge White said he is also inclined to sanction Russo and require him to pay Support Community roughly $18,000 in legal fees incurred for the time its counsel, Gary S. Sedlik of SedlikGroup PC, spent identifying the errors and litigating the AI issue.
The judges added that he may require Russo to self-report any final sanctions ruling to the California State Bar and require Russo and everyone at his firm to complete at least three CRE hours on the ethics of AI use.
In the motion to withdraw, Russo said he took "full responsibility for these citation errors," but he explained that he had been recovering from COVID-19 during the "critical brief preparation and finalization phase," and he had relied on "research assistance," which, upon review, included AI-generated errors that were not properly verified.
During the hearing on Friday, Russo noted that the legal industry has embraced AI tools with verification systems and he has already completed over 15 CLE hours on the topic in light of the judge's tentative ruling.
Russo also said he regrets wasting the court's time on the issue, but he argued that if he is required to report any sanctions to the state bar in the Golden State, he will likely have to report the matter to other state bars in other states where he practices and works pro bono. He argued that such a punishment for something that is becoming more commonplace in the legal industry could have a "domino effect" and deter future pro bono work.
"As I've gotten older, I've spent more and more time doing pro bono service for a variety of courts and it's kind of an aberration to have this particular situation that occurred when I was suffering under Covid to give rise to an outcome where my plans to do more and more pro bono work for those courts [impacted] by a black mark against my record," he said.
Russo also disputed the plaintiff's argument that he used AI to produce the motion in its entirety, and he noted that he had offered to donate $20,000 to the plaintiff's employer, Ronald McDonald House, to rectify the issue and void "satellite litigation" over it, but the plaintiff's counsel quickly rejected the offer. Russo additionally criticized Sedlik's $18,000 fee estimate, and he argued that it raises questions about his billing practices.
"I'm not blaming him for anything," Russo added. "I am saying that at some point he said, 'Look, you're going to have to pay me $40,000 in legal fees if you want me to go along with the stipulation to withdraw the brief,' which I found to be frankly very insulting."
But Sedlik argued that the error-filled filing constitutes a "fraud on the court," and although AI-generated errors in filings have become an "endemic problem now in state and federal courts throughout the country," Judge White should still issue tough sanctions against defense counsel in this case.
Sedlik pointed out that the motion at issue contained roughly 12 to 14 citation errors — not just one or two — including fake cases and extensive quotes that do not exist, thereby suggesting that no one reviewed the AI-generated motion before filing it. And although Sedlik emphasized that he has no personal animosity against opposing counsel, Sedlik argued that Russo's explanations for the mistakes have evolved over time and ultimately amount to illegitimate excuses.
"The excuses range from that he was sick, to he had a bee sting, to his associate was sick and his paralegal had issues, which we don't need to get into today, but to me those don't sound like any justification for what happened here," Sedlik said.
Sedlik also pointed out that after he informed defense counsel of the errors, Russo and Sedlik had numerous exchanges, and Russo refused to pay Sedlik's client his legal fees up front for having to deal with the issue, and instead blamed him for using AI to catch the citation errors, which Sedlik said was "a little bit offensive."
"Why should my client have to pay one penny because of what Mr. Russo did here?" Sedlik said. "It's a question that I can't answer and I don't think Mr. Russo has an answer now."
Judge White pointed out, however, that Sedlik did not catch the AI-generated errors until after briefing on the summary judgment motion was complete and just before a hearing on the motion, so the judge doubted that he should award plaintiff's counsel fees for opposing the motion.
But Sedlik replied that the errors at issue were in sections of the summary judgment motion that his client did not need to respond to, since the contract dispute is largely fact-based and does not involve substantive arguments over the relevant legal authority. He added that it is not his job to check all of Russo's citations.
"That's not something obviously that I have a legal obligation to do, or I would normally do frankly when I'm responding to a motion, even a summary judgment motion," he said.
As further support, Sedlik cited the recent California appeals court published ruling in Noland v. Land of the Free
in which a panel imposed a $10,000 monetary sanction against an attorney, and wrote that it should serve "as a warning" to Golden State attorneys to personally review case law quotations made by generative AI.Sedlik additionally defended his firm's hourly rates, noting that he has 30 years of experience, and they represent a third of what his former colleagues at an international law firm charge.
At the end of the hearing, the judge took the arguments under submission, and he did not rule from the bench. Throughout oral arguments, Judge White acknowledged that he does not know much about the specifics of the AI technology at issue, but he said he takes the matter seriously. In his tentative ruling, he also noted that this is the first time he has come across this issue.
In recent months, the potential problems with using generative AI tools to draft legal filings have created havoc in litigation across the country, leading to sanctions, attorney disqualification and referrals for disciplinary proceedings.
Those headline-grabbing rulings came down in district court cases that include Mavy v. Commissioner of Social Security Administration
in Arizona, Johnson v. Dunn in Alabama and Mata v. Avianca Inc. in New York.Meanwhile, in California state appellate court earlier this month, a Boies Schiller Flexner LLP partner representing women who allege the Church of Scientology harassed them for reporting convicted actor Daniel Masterson's sexual assaults asked a panel to strike a brief containing AI-generated citation errors, saying he "very much regrets" the errors, but they should not impact his clients' case.
Support Community was represented during the hearing by Gary S. Sedlik of SedlikGroup OC.
MPH International was represented during the hearing by Jack Russo of Computerlaw Group LLP.
The case is Support Community Inc. v. MPH International LLC, case number 4:23-cv-04911, in the U.S. District Court for the Northern District of California.
--Editing by Vaqas Asghar.
For a reprint of this article, please contact reprints@law360.com.
