By John Byrne
Let this be your daily reminder not to use fake cases in your briefs. The Eleventh Circuit devoted an entire section of its recent opinion in Akerlund et al. v. Atlas Air, Inc. to lambasting a lawyer for citing hallucinated cases in his brief even after he was warned. Opening that section by noting that it had “saved the worst for last,” the Court got into granular detail about the lawyer’s misuse of AI. Judge Grant wrote the opinion, joined by Judge Rosenbaum and Judge Kidd. The Court wrote:
"We expect that when lawyers submit briefs, they give us their best view of what the law is, and how that law supports their clients. We read those briefs carefully—not just because they are the product of counsel’s time, effort, and skill, but because they help us as we try to reach the right answer. All that is lost if a lawyer decides that it is not worthwhile to do the work to persuade us—if he is willing to sign his name on whatever a machine churns out without so much as checking it for accuracy. We cannot do our job the same way unless lawyers do theirs.” You can read the opinion here.
12 comments:
Two thoughts - First, I wish it didn't take hallucinated cases in a court filing for courts to call out--in a very public fashion like in this opinion--a lawyer's lack of competence. A large portion of our bar could use this kind of public shaming, whether they use AI or not. Second, but related to competence, HOW IS THIS STILL HAPPENING??? There have been so many other public shamings for hallucinated cases in the past 5 years. Yes, this was an appeal so, presumably, the case may have started in the district court before these shamings were as widespread. But as the Court makes clear, these hallucinations were in Sabatini's appellate brief. By then, he knew better.
There really needs to be a new parenthetical or citation convention when citing a hallucinated case. Footnote 5 is an invitation -- or at least breadcrumbs for a future data-scraping hallucinator -- to cite a fake case, perpetuated by the court through no fault of the court.
The court writes that it "read[s] those briefs carefully—not just because they are the product of counsel’s time, effort, and skill, but because they help us as we try to reach the right answer." You would think the primary reason they would read them is because we have a party presentation system and it is their job. But maybe the implication, if you read the opinion carefully, is the imperious judge need only glance at materials but may deign to truly consider them provided there is some basis to think it is worth their effort. I know it was a call to be better, but leaving out our adversarial and party presentation requirements (particularly a court so quick to find waiver) seemed a bit sloppy to me.
Sabatini is a joke and should be severely sanctioned by the bar. But probably won’t be because they don’t want to upset the right. Always filing complete BS right wing political lawsuits. No regard for the law and no belief in the constitution or the oath of attorneys.
I respectfully dissent. 1 AI is here. Like it or not. 2- AI is a valuable tool when used properly. 3 - A lawyer filing a brief with AI fake cases is as lazy as a lawyer filing a brief ridden with typos, farmed out to off shore paralegals with poor reasoning and quotes not properly attributed. The answer here is to read what you write and edit edit EDIT. AI will get better and it’s not going away. Just like using word editor or on line grammar editing tools (many marketed to lawyers for briefs a few years ago before widespread use of AI) the solution is not attacking AI and its use but lawyers being responsible for their work product. AI tip- when asking it to review a document or research an issue use this phrase in the request “do not make up any cite case, rule or fact to please me. All the work you return must be 100% accurate and verifiable “ and then after you receive something, review it CHECK EVERY CITE, and send it back to AI to review with the same warnings. AI can review a 300 page transcript and highlight every mention of a relevant word - like gun, kilo, etc. it can quickly summarize medical records or thousands of pages of bank documents to locate relevant transactions. Like the introduction of computers in the late 1960s it breeds fear based on ignorance. For a good sense of how computers were viewed and feared in the late 60’s the last season of Mad Men has an IBM mainframe being installed and characters freaking out. And yet today we cannot do most anything without computers. So the solution here is not to attack a tool but to warn that the tool must be properly supervised and used correctly. For those reasons I respectfully dissent. H Rumpole, Esq. State Blog proprietor, and misanthropic gadfly.
Well-written and I know you’ll disagree with me but your premise is based on “And yet today we cannot do most anything without computers” as if that’s better. It just isn’t. Of course one would have to be old like me to know or argue that point. Life before computers was quite different for sure, but a regular day back then was like what we consider a vacation day now.
Your dissent is a total non sequitur. No one is attacking AI. Everyone (with critical thinking skills) knows it's not going anywhere. Of course it will get better. And the entire tenor of section 5 of this opinion is exactly as you say--if you're going to use it, "edit edit EDIT" and "CHECK EVERY CITE." If anything, your comment is a concurrence.
FYI state judges were given access to AI and dozens and dozens of them have their own AI accounts they use in writing orders. And don't tell me those clerks in Atlanta are not using AI to summarize briefs for the bench briefs they give to the Judges. I sincerely doubt a circuit judge read this brief and caught the errors. One of their clerks did. So now what? So much for each brief being carefully scrutinized. They get a brief on a 4th amendment search issue and I'm sure the chatter is "don't they know the court they filed this in? Come on--- who wants to write the PCA with a string cite? Send it to Judge so and so's new clerk. Let her write it." That's how things are really done in Atlanta.
This is a fascinating (and super important) point - are you saying that by including the fake case names and fake reporter citations in their REAL opinion, the 11th is inadvertently making this problem worse??? Because now AI can scrape fake cases and reporter positions from a REAL opinion and won't know they were "cited" only to show they weren't real???? That's really important for judges to know, if your read on AI algorithms is correct.
@10:39, it only "makes the problem worse" for lazy and unethical attorneys. All of us who do our job correctly read every case before including a citation in a brief. If AI generates a cite based on reference to a hallucinated case in an opinion like this -- the only way that cite makes it into a filed brief is if the attorney fails to do that basic fundamental thing. Anyone who cites a case without reading it deserves whatever comes from it.
@11:54 - Agree. But since we know there will always be (as there has always been) lazy and unethical attorneys, it just makes the problem worse.
Sabatini is such a chud
9:18 is right
Post a Comment