An AI Screw-Up By... Sullivan & Cromwell?
There are multiple ironies involved in this unfortunate incident.
Welcome to Original Jurisdiction, the latest legal publication by me, David Lat. You can learn more about Original Jurisdiction by reading its About page, and you can email me at davidlat@substack.com. This is a reader-supported publication; you can subscribe by clicking here.
Tonight brings huge news out of Sullivan & Cromwell.1 We can now add S&C, truly one of the world’s most prestigious and profitable firms, to the AI “Hall of Shame”: the list of law firms that have submitted court filings containing AI-generated “hallucinations.”2
On Saturday, April 18, Andrew Dietderich, founder and co-head of SullCrom’s global restructuring group, sent a letter to Chief Judge Martin Glenn of U.S. Bankruptcy Court (S.D.N.Y.). In his missive, Dietderich apologized profusely for a motion filed by S&C in a bankruptcy case that contained “inaccurate citations and other errors,” which he helpfully listed for Chief Judge Glenn in a three-page attachment (because there were that many—the three pages were single-spaced).
As you’d expect, this AI screw-up by S&C has been covered by, well, everyone—including Above the Law, Bloomberg Law, Business Insider, Law360, and Reuters. I don’t have much to add to the copious coverage; I’ll just point out two ironies.
First, as noted by Joe Patrice of Above the Law, Sullivan & Cromwell is the law firm that… advises OpenAI on the “safe and ethical deployment” of artificial intelligence (a representation S&C touts on its website). Physician, heal thyself?
Second, as Dietderich explained in his letter to Chief Judge Glenn, the AI errors were called to his attention by… Boies Schiller Flexner. Why is it ironic that BSF, one of the nation’s leading litigation firms, caught these mistakes? Because until now, Boies Schiller was the reigning champion of “Elite Law Firms Guilty of AI Fails.” (In addition to the AI snafu chronicled in that story, which took place last year, BSF subsequently failed to catch AI-generated errors in a filing by co-counsel in a different case.)
When I covered BSF’s AI error last year, I gave props to then-partner John Kucera: he “fully and freely admitted the errors, and he didn’t throw any colleagues under the bus.” I’ll say the same here about Andy Dietderich. He’s a senior partner at S&C, having joined the firm almost 30 years ago; he leads its restructuring practice, which he founded; and he’s a giant of the bankruptcy bar, Chambers Band 1. Nobody was paying him $2,500 an hour to do legal research; this screw-up was clearly the handiwork of an associate (or perhaps I should say former associate). But Dietderich signed the letter solo, and he didn’t point the finger at anyone else.
So that’s my compliment for S&C. Here’s my criticism: considering how much time has passed since the earliest AI errors—Avianca-gate took place three years ago next month—mistakes are less and less excusable (to the extent that they ever were). After all, as noted by Joe Patrice, there are tech tools out there—some of them powered by AI (a third irony?)—that are designed to catch AI hallucinations. Patrice gave a shoutout to RealityCheck by BriefCatch (and as a board member of BriefCatch, I feel it’s my fiduciary duty to include shameless plugs for the company whenever I can).
I’ll conclude my complaints about S&C here—because people in glass houses shouldn’t throw stones. As some of you might recall, earlier this year I had an AI oopsie of my own (see the update to my SCOTUS tariffs ruling write-up). I’ll simply repeat my two signature quips, urging you to extend grace when it comes to AI fails:
“In the future, everyone will be world-famous for 15 minutes… for an AI screw-up.”
“The next time you hear about an epic AI fail, instead of (or at least after) laughing your ass off, perhaps have the humility to say this to yourself: ‘There but for the grace of God go (A)I.’”
Or, if you prefer, here are some bon mots from Claude, which it generated after I fed it my two sayings and asked for more along the same lines:
“An AI making things up with total confidence isn’t a bug. It’s a mirror.”
“An AI confidently gives wrong answers. A human confidently gives wrong answers. One of them gets a performance review.” (And only one of them gets fired from a $225,000-a-year job.)
“Garbage in, garbage out—but now the garbage speaks in complete sentences and cites its sources.”
Not bad, eh? As I keep telling my husband Zach, who views AI with deep skepticism, it’s actually great—as long as you use it responsibly.
And no, it’s not the departure of appellate and Supreme Court superstars Jeffrey Wall and Morgan Ratner, who will be leaving S&C for Gibson Dunn. Reported earlier this evening by C. Ryan Barber and Erin Mulvaney of The Wall Street Journal (gift link), it’s the early frontrunner for Move of the Week. Wall served in the first Trump administration as acting solicitor general, and Morgan Ratner is also an alum of the Office of the Solicitor General (as well as a former OJ podcast guest).
[UPDATE (7:16 p.m.): Gibson Dunn’s PR firm drew my attention to the firm’s press release, which notes that four S&C partners are joining GDC. In addition to Wall and Ratner, Judson (Judd) Littleton and Yaira Dubin are also making the move. All four are former SCOTUS clerks who spent time at OSG (Littleton as a Bristow Fellow).]
If it’s any comfort to S&C—and I realize it probably isn’t—legal technologist Damien Charlotin keeps a running tally of cases where generative AI produced hallucinated content, and so far he’s uncovered more than 1,300 examples.
Thanks for reading Original Jurisdiction, and thanks to my paid subscribers for making this publication possible. Subscribers get (1) access to Judicial Notice, my time-saving weekly roundup of the most notable news in the legal world; (2) additional stories reserved for paid subscribers; (3) transcripts of podcast interviews; and (4) the ability to comment on posts. You can email me at davidlat@substack.com with questions or comments, and you can share this post or subscribe using the buttons below.
Apologies, readers—I wrote this under time pressure (because I had to start watching the kids when our child care timed out), and the original version of this post contained a number of typographical and grammatical errors (despite my proofreading it multiple times and running spellcheck in Word).
I believe I've caught everything. But if you see anything I might have missed, please feel free to let me know here in the comments (or email me if you prefer). Thanks!
It sounds as though the "RealityCheck" product might be one of those guns that kicks as hard as it shoots — that is, it may be a product with powerful defensive AND OFFENSIVE uses.
When such a product exists, in at least some circumstances, the relevant standard of professional care might oblige a lawyer to check not only the accuracy of his own citations, but those of his opposing counsel.
Not all AI hallucinations are consequential: Some of the made-up citations and holdings embody or reflect real citations and holdings that the AI has turned into ... I dunno, a blended-average hallucination. If you catch your opponent citing a bogus case but for an uncontroverted and incontrovertible principle of law, do you bring it to your opponent's attention? Or to the court's? I'm inclined to think the former, as a matter of professional courtesy and personal grace.
But there are certainly cut-and-slash opponents who'll let pass no opportunity to do harm to you or anyone on your side.
This profession is getting scarier, even if you're keeping up with technology. If you aren't, it's rapidly becoming an incomprehensible profession.