Flycatcher v. Affable Avenue, LLC. was a case in the Southern District of New York in which hallucinated citations from a large language model ended in sanctions for the offending attorney.
Affable Avenue, represented by Steven Feldman, was among over a dozen defendants in the suit brought by Flycatcher. After Feldman entered a motion to dismiss, counsel for co-defendant Top Experience Company (“TEC”) reviewed the citations. He found that 13 of the decisions did not exist and Feldman had misrepresented the holdings of eight others. He alerted Feldman to the mistakes and gave him four days to retract the filing, but Feldman did not act.
TEC’s counsel then filed a whistleblower letter to alert the court to what his co-defendant’s counsel was doing.
The court found that not only were 13 citations hallucinated, but the writing style of the memo of law was inconsistent with Feldman’s other filings to the point of being suspicious. The judge concluded that Feldman had used a large language model to write the motion — which, the judge noted, allowed the attorney to misstate the law, a breach of court rules . The court issued an order to show cause to Feldman to explain his actions.
In a second filing, in which Feldman requested that the court allow him to re-file a revised brief, Feldman raised eyebrows with his “convoluted” response and some “bizarre” prose that misquoted Ray Bradbury’s Fahrenheit 451 and referenced ancient Egyptian texts. In other places, Feldman explained that Westlaw and Lexis were prohibitively expensive, and that he was forced to use free tools like Google Scholar.
Plaintiff Flycatcher moved for sanctions, and in opposition to the motion, Feldman filed a reply that contrasted greatly with his AI-generated filings, full of grammatical errors and run-on sentences. At oral argument, Feldman gave a convoluted explanation about using tools to find cases and other tools to verify them.
Feldman filed a letter with a third document (without the court’s permission): a proposed revised memo of law to oppose sanctions. Unfortunately, even that filing cited to hallucinated cases, now a third time.
In its opinion and order, the court emphasized that using AI tools to write briefs in itself did not constitute bad faith. Rather, sanctions were warranted because Feldman chose to continue to engage, on two more occasions, in filing faulty or non-existent citations. He had not learned from his mistakes.
Ultimately the court sanctioned Feldman for filing with the court three documents with citations to cases that didn’t exist. The court struck his answer, which in effect gave the plaintiff a finding of liability against Feldman’s client, Affable Avenue. Additionally, TEC’s counsel was awarded attorneys fees for being forced into being a whistleblower.
The moral of this story is nothing new: We are obligated as attorneys to read the cases that we file with the court. Perhaps relying on associate attorneys or co-counsel who have reported that they read the entire case is reasonable, but it is still paramount for the person signing to know that the law they are relying on, and that they’re asking the court to rely on, is real.
Basic intellectual honesty and owning our mistakes has always been a rule, and AI has not changed that. Maybe AI undermines each of our obligations to be the best attorneys we can be because it offers the answers more quickly. Indeed, Feldman said he didn’t have time to citecheck what he was turning in to the court. I would submit that we all — lawyers and clients alike — need to make the time to know that what we’re submitting, and what we’re saying, is true.
