The Volokh Conspiracy
Mostly law professors | Sometimes contrarian | Often libertarian | Always independent
Argument Against Sanctions for Lawyer's Filing of Motion That Contained AI-Hallucinated Cases
The memorandum of law, filed by the lawyer's counsel, is here. It seems quite well argued, though I don't agree with the backup argument (in Part II.E) that the lawyer's reliance on ChatGPT as a source for the supposed text of precedents was not just innocent but "reasonabl[e]." Here's the opening paragraph of the introduction:
In the Order, the Court describes this situation as "unprecedented." We agree. We can find no case where, as here, a lawyer using a new, highly-touted research tool obtained cases that the research tool itself completely made up. The lawyer, Mr. Schwartz, had no idea this was happening, even when opposing counsel brought their inability to locate the cases to his attention. ChatGPT even assured him the cases were real and could be found on Westlaw and LexisNexis, and continued to provide extended excerpts and favorable quotations. Now that Mr. Schwartz and the Firm know ChatGPT was simply making up cases, they are truly mortified; they had no intention of defrauding the Court, and the mere accusation – repeated in hundreds (if not thousands) of articles and online posts – has irreparably damaged their reputations. They have apologized to the Court in earlier submissions and do so again here.
Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Comments may only be edited within 5 minutes of posting. Report abuses.
Please
to post comments
Innocent? Sure, I'll believe that, because it was sure to be caught, so there's no way they'd have done it deliberately.
Reasonable? Not in the slightest. The fact that ChatGPT makes stuff up is widely discussed, and they had plenty of red flags go up, they admit as much themselves.
And asking someone if they are lying is a well-known trick question. Any lawyer who hasn't heard that in 30 years of practicing isn't ready for Carnegie Hall.
Brett, you're terminally online. The vast majority of people are not. If your source of knowledge is that lots of people online are talking about something, consider that this is not an indicator that it's well known.
As I said in the other thread, the Stupid Defense. (Which often has the great advantage of being true!)
Sometimes the best argument you have is to just fall on your sword.
No, falling on your sword would be saying "Yes, court, I should be sanctioned. I screwed up, I apologize, and a monetary penalty is completely appropriate."
He doubled down on stupid:
https://twitter.com/innercitypress/status/1666851080389505043?s=20
In short — as I noted the other day — he really doesn't know anything at all about federal practice.
Got to give him credit for providing so much comedy, though. Wonder if the judge started laughing. I would.
Judge Castel: Have you heard of the Federal Reporter?
Schwartz: Yes.
Judge Castel: That’s a book, right?
Somewhat off-topic, but in the age of digital databases, the West reporter system seems to have outlived its usefulness. I know that TV shows like to show law offices with lots of books, but the reality is that 99& of research nowadays is done online. In the last, say, 10 years, I think I only went to a physical law library three times, and each was to look for obscure sources, not the mainstream caselaw.
The last time I worked at a firm that had a set of reporters was roughly 20 years ago. When that firm shut down, it liquidated its assets, and it couldn't find anyone willing to take the books. The amount of space they took up versus their continued utility…
Indeed, aesthetics were pretty much their only utility. (I'd have taken them myself because I think throwing out books is a sin, and I also liked the aesthetics, but I just didn't have any place for them.)
Mistakes happen, and I'm inclined to be forgiving of them if they're owned up to. The problem with this particular mistake is that the other side had to spend a lot of money in attorney's fees as a result.
If I were the judge, I'd strike the brief that had the made up cases, give the plaintiff 20 days to file an amended brief, and order the plaintiff to reimburse the defendant for its attorney's fees that were incurred as a result. While at the same time making it clear to the defendant that I expected them not to gouge the plaintiff in determining its attorney's fees.
Did they? It looks their only filings since the submission of the fake cites is a 6 page reply to the opposition that cited the cases, and a 1 page letter after they filed "copies" of the fake cases pointing out that they were, in fact, fake.
I didn't check the docket, but assuming you're right, it doesn't change my underlying point: Make the other side whole but don't give it a windfall.
This isn't really about the other side — which I'm guessing would be perfectly happy with a dismissal; they don't need a check for a few thousand dollars — but about the integrity of the court.
To me, it wasn't done maliciously so the client's case shouldn't be sanctioned. But the lawyer who did it should absolutely be sanctioned. I think he should be held in contempt and/or disciplined by the local disciplinary authority. It's not about the case, it's about the lawyer's conduct in a courtroom.
I would not be surprised to see these guys hit with a bar referral, too. There are several issues of competence, and the "local counsel" didn't read any of the pleadings.
The problem with this argument is that, as one of the key cases they cite indicates, "conscious avoidance may be the equivalent of knowledge." Citing cases without looking them up, and failing to take any reasonable measures to verify their existence after the court and opposing counsel have indicated a high likelihood that they're made up, seem like it typifies "willful ignorance".
As I say above, I don't think this was in any sense deliberate, because if he'd known the cites were fake, he'd have known there wasn't any way he'd get away with it.
But he did press on regardless of numerous red flags, and why? Because it was too much of a bother to actually check the cites? Really?
I think the appropriate sanction would be to have the dude refund any money he has already gotten from his client, and cover the expense of his client hiring competent counsel, and that's how it should be pointedly phrased.
What I would do is remind LEXIS was once new, and that I didn't first trust it when it arrived, but even the Court now trusts it -- and I made the mistake that AI was just a new thing like LEXIS, that I didn't know how it worked, but only that it did. But it doesn't and I apologize for my mistake.
If you used Lexis when it first came out without bothering to actually determine it was a real legal search engine instead of monkeys at a typewriter, then that would be a problem too. You can't just trust a brand-new program without having done anything to verify that it's a reputable source. "I heard something about it from my kids" is not adequate.
If a judge orders an attorney to submit copies of cited cases, it seems reasonable to expect the attorney to locate and submit a copy of the opinion, not a paragraph from the case (which is what GPT provided). If all you can find from the source you're using is a paragraph, and not the full opinion, that's a pretty big red flag. Also, Lexis, WestLaw, and Fastcase are not the only options for looking-up online (and verifying) case citations. Someone needs to inform the attorney that there are free websites (like Courtlistener) where you can search judicial opinions using a variety of different criteria (like the citation).
There's Google Scholar. There's asking a lawyer at another firm that does have those services. There's (gasp) going to a law library. But this guy isn't the sharpest light bulb in the drawer.
But this guy isn’t the sharpest light bulb in the drawer.
Bet you he mixes metaphors, too.
Does the bear wear a funny hat?
(Could re-mix that metaphor, but it's a bit soon after yesterday's surgery.)
Interesting memo. So it wasn't crutching on the technology to do the lawyering, but trying to get access to the databases that he needed, and that he thought he had to pay for at WestLex. Essentially using it as a paywall workaround. He was gaming the system.
This comes from a certain way of thinking about the law--as an objective repository of principles, one which just has to be sorted and filtered the right way in order to provide the right answer in any given case. Manually Sheperdizing a few precedents used to be enough to disabuse folks of that notion; now they're just slaves to the traffic lights in the headers.
O tempore, (billed in 15 minute increments), &c, &c.
Mr. D.
Assuming that subjective bad faith is required for a trial court to impose sanctions sua sponte, I have difficulty understanding how any experienced attorney could have believed (even subjectively) that citing purported authorities, which he had not read, to a federal district court constitutes ¨an inquiry reasonable under the circumstances¨ within the meaning of Fed.R.Civ.P. 11(b).
Newsflash: There are really bad, really stupid lawyers out there.
Think of this as roughly analogous to a lawyer never having used Westlaw before who — frantic to respond to a motion he doesn't understand — types up a natural language search, gets a case spit back at him, and quotes what sounds like a great excerpt from it, without understanding the significance of the red flag at the top of the screen. ("I just assumed it meant the case was really important.")
"We can find no case where, as here, a lawyer using a new, highly-touted research tool obtained cases that the research tool itself completely made up."
It starts off by assuming the answer. This isn't about a lawyer using a genuine research tool. Chat GPT is not a research tool. It's a text generator. The only reason the lawyer gave in his statement for believing it was a research tool was he assumed it from what he'd heard from his kids. He missed doing the first step, which was verifying that he was using an actual "highly touted" research tool.