I’ve lamented before that it seems like I could spend every day writing about another example of lawyers not learning the obvious lesson that using generative AI in connection with legal research or writing court filings is more trouble than it is worth.
Even if the toll that GAI is taken on the planet weren’t overwhelming, it is more trouble than it is worth just based on the overwhelming evidence that the “generative” part of it is always going to result in risk of feeding you false information.
Just in the last 20 days or so, all of the following stories have shown up in legal media.
Third Circuit finds AI hallucinations involved lack of competent representation.
Attorney sanctioned for hallucinations in workers compensations case
Federal judge in Utah sanctions civil rights attorneys who admitted to AI hallucinations
Attorneys representing City of New Orleans sanctioned for AI hallucinations
Oregon lawyer (different one) sanctioned $10,000 for faked case citations and quotations
DC lawyer admonished by Fourth Circuit for citations to nonexistent cases
That’s 9 in fewer than 3 weeks and those are just the ones that have made it to court rulings and been picked up by either Law360 or Bloomberg Law. And none of those are disciplinary proceedings which may or may not also follow after the imposition of sanctions.
Why I am prompted to write though (pun most certainly intended) is that Tennessee now has its first instance of public discipline being imposed on a lawyer for the filing in a tribunal of AI slop.
On March 30, an attorney in Dyersburg, Tennessee [a city about 90 minutes by car from Memphis and with a population of around 20,000] received a public censure in connection with conduct involving filing a reply brief in a matter pending before the Armed Services Board of Contract Appeals. After that tribunal reviewed the filing, it found that more than 70% of the citations in it were fictitious. The lawyer’s defense was that he believed a paralegal in his office had reviewed all of the citations before filing. You can read the press release about this discipline below:
Remarkably, the public censure only found violations of RPC 1.3, 5.3, and 8.4(c) and did not include any finding of a violation of RPC 3.3. If the Armed Services Board of Contract Appeals is a tribunal, then it is surprising that there was no violation of RPC 3.3(a)(1) included. Even if that were a non-adjudicative proceeding, then there is the issue of compliance with RPC 3.9.
That being said, public censures are often the product of negotiations so it could simply be that the rules referenced are the only ones on which a deal acceptable to all were struck. It also is not known whether, for this lawyer, this public censure is the first discipline of any type received or if the reason this amounted to a public censure what the lawyer had already had private discipline in their history.
It is also worth noting that since the tribunal did not say “cases” but rather “citations,” that the hallucinations underpinning this discipline could well have involved citations other than case law, such as citations to parts of the record and contents of transcripts.
That fact is worth noting because, and I know I repeat myself, generative artificial intelligence does not exist for the purpose of providing users with accurate information. It exists for the purpose of creating new works based on its training data and other information to which it has access.
My co-author of Professional Responsibility in Litigation and friend, Doug Richmond, is often appropriately cited and credited for telling lawyers for years that the “e” in e-mail stands for evidence.
Someday lawyers are going to figure out that the “g” in GAI stands for “generating problems.”