That revelation in a TED talk from Katyal generated some criticism from those who questioned the wisdom of owning up to using AI in Supreme Court advocacy, but it also illustrates how useful AI tools can be in preparing attorneys for court, according to scholars and legal technology professionals.
AI can help lawyers find patterns in huge amounts of data, like Supreme Court opinions, which is how Katyal said he used it. And it can be employed as a digital moot court, allowing lawyers without the resources for elaborate rehearsals to practice their arguments.
Despite the criticism of Katyal's talk, the truth is that many lawyers and advocates are already using the technology this way, even if they haven't publicly admitted it, scholars add.
"He just said the quiet part out loud," noted Mitu Gulati, a professor at the University of Virginia School of Law.
Still, given the growing track record of AI tools introducing hallucinations into briefing, scholars warned that lawyers need to be careful about their use of the emerging technology — particularly when they're using it to get ready to appear before a judge.
The Criticism
In his April TED talk, Katyal described the four teachers he worked with to prepare for November's oral arguments over the legality of President Donald Trump's tariffs and to "connect" with the justices, who ultimately ruled in his favor.
One was an athletic coach, one an improv coach, one a meditation coach and another was Harvey, the legal-focused artificial intelligence tool.
Katyal said he's been building his own "bespoke system" with the company Harvey AI for the past year.
"I trained it on every question asked by a Supreme Court justice in the last 25 years and everything they've written, every opinion, every concurrence, every dissent, every separate opinion. And in that, patterns emerged," Katyal said.
He described how Harvey foresaw Justice Neil Gorsuch's question about taxing power, Justice Brett Kavanaugh's queries about the difference between tariffs and embargoes, and Justice Amy Coney Barrett's concerns about tariff refunds.
That admission earned Katyal a lot of criticism from other attorneys and court watchers.
"Did Neal think this through? How does he think the justice will react to seeing her question compared to the AI output?" South Texas College of Law professor Josh Blackman asked in Reason magazine about Katyal's claim that Harvey predicted Justice Barrett's questions. "Even if the AI anticipated her question, what possible sense does it make to advertise this point to the world?"
Katyal, Harvey AI and the Supreme Court did not respond to interview requests.
The negative reaction to Katyal's talk doesn't surprise scholars who spoke with Law360 Pulse.
Lawyers aren't generally known as early adopters of technology, said Kevin T. Frazier, AI innovation and law fellow at the University of Texas School of Law, who called many attorneys' attitude toward AI use "unfortunate and unsustainable."
"A lot of those criticisms come from a place of fear and a lack of understanding," echoed Nicola Shaver, co-founder and CEO of Legaltech Hub.
That fear may stem from the fact that AI potentially makes lawyering "less special," Gulati said.
"There is this model of lawyering — the priest in the temple — which is that we have a special language and we know what it means when the justice twitches their nose," Gulati said. "AI is democratizing that. That's the source of at least some of the resentment."
Despite the negative reactions to Katyal's admission that he used AI to prepare for oral arguments, though, many attorneys may quietly see its value.
"A lot of lawyers and litigators and law professors are using these new tools and technologies, but usually they're not as up front about it," said Xiao Wang, director of the Supreme Court Litigation Clinic at the University of Virginia School of Law. "In some ways, the legal profession is very tradition-based and very slow moving to adopt certain technologies. So it might be the case that a lot of lawyers don't want to openly say, 'Hey I've been using AI.'"
"But they can probably see from Neal's talk and from their own experience that there are definitely some useful applications of the technology," Wang told Law360 Pulse.
Leveling the Playing Field
AI can be especially useful in helping lawyers prepare for oral arguments, at the Supreme Court and in the lower courts.
That's easier to do when arguing at the Supreme Court, where lawyers always face the same nine justices, Wang acknowledged. It's more difficult in the lower courts, which have more judges and where the judges who hear arguments vary.
AI can still provide a digital version of moot court for attorneys arguing in the lower courts, who often don't have much oral advocacy experience and rarely get to practice before a moot court, he said.
"It's going to be really helpful if you can't afford to hire associates that clerked for every judge and carve out time for that, to have a resource like a large language model," Wang added. "I think that can be a really powerful tool to level the playing field."
There are even AI moot court platforms that are specifically designed to enable lawyers to engage in courtroom argument preparation, according to Shaver.
Frazier said judges and clerks can also use these tools to identify questions for attorneys, leading to "richer argument that more narrowly addresses the most important issues."
An AI tool could even help attorneys avoid the courtroom completely if it predicts a poor outcome to arguments, Gulati pointed out.
"If I can feed in my Second Circuit panel's opinions and I can predict using AI that it looks like I'm going to lose 2-1, I'll settle the case based on that," Gulati said.
Exercise Judgment and Verify
That doesn't mean there aren't potential pitfalls when attorneys use AI to prepare for oral arguments, scholars warn.
A ballooning number of attorneys have already been sanctioned or criticized by judges for filing pleadings that include fake legal citations conjured by AI hallucinations.
Those hallucinations can just as easily show up in oral arguments, according to scholars. So lawyers should make sure they exercise their own judgment, review all AI suggestions and verify any citations and case law, they say.
Attorneys should also use legal-specific AI applications that have access to a sound body of legal content, according to Shaver, who said that purely legal models are much less likely to hallucinate than more generic platforms that draw on all the information on the internet.
"When you get up in front of the judge, you should be able to stand behind every proposition, every interpretation, every argument you're making, as well as every citation and every quote within your submissions and in what you've provided to the court," Shaver said. "If you cannot do that, it's not a failing of the AI, it's a failing of your obligations as a lawyer."
Even when AI doesn't conjure false cases and information, it can spit out answers that are correct but superficial, without the level of nuance or depth an attorney needs during arguments, Wang pointed out. So he said attorneys shouldn't settle for an AI response to their first question, but should ask follow-up questions to get deeper answers.
"The lawyers that fail to lean into this AI moment," Frazier said, "will find themselves at a growing disadvantage and will do a disservice to their clients."
--Editing by Alex Hubbard.
For a reprint of this article, please contact reprints@law360.com.
