Hi, it’s me again. I really don’t want to be writing about lawyers continuing to get sanctioned nearly daily at this point it seems for filing things with the court containing AI hallucinations. I know I always have excuses for these long stretches where I don’t post anything to the blog.
This time (my May/June disappearance) has more to do with other circumstances, but a part of it has been that every time I sit down to write about something else (anything else) I’ve just seen yet another instance of such a problem, or someone arguing about how the next step that is missing is just some other software or AI product that can check the work of the first AI product to protect lawyers from these kinds of situations. These are usually also read in close temporal proximity to someone continuing to insist that the only answer to the rise of Generative AI is that lawyers have to make sure that they get on board with using it for as many pieces of the practice of law as possible.
So, I repeat myself but again would suggest a different action plan. . . how about we don’t do that? How about, instead, we start having a conversation about what should actually happen? What should actually happen is that lawyers should stop using lie machines to do their legal research and draft their court filings.
And, yes, “lie machines” sounds inflammatory, but it captures a pretty good sense of what the “generative” part of Generative AI is if you use it for doing your legal research or generating your court filings where it is going to be pulling content for you from transcripts or recordings or exhibits, etc.
The tool lawyers are using is seeking to create new works based on its training set (whether or not you are working with a tool that also is able to retrieve reliable sources before creating that new work). Inherent in that is that lawyers are using a machine that is, at some point, going to produce results for you that are going to be lies. Using such a tool is a shitty foundation for reliable lawyering. That would be true even if it were not also unsustainable in terms of energy usage and environmental impact.
But, of course, these tools also require energy levels and water consumption that are unsustainable.
And, in case you missed them and want them for your reading pile, here’s a subset of all of the instances that have come to light just in May and June 2026 of lawyers getting hit with the consequences flowing from continuing to use the lie machine.
Among the most significant individual sanctions on a lawyer happened in May 2026 flowing from using generative AI for research and brief writing is a six-month suspension order issued in Georgia. The setting involved a criminal case in which a defendant had been sentenced to life in prison. An Assistant District Attorney filed something in response to a motion for new trial that resulted in the trial court entering an order denying a new trial that included fake case citations. On appeal, the same ADA filed more items with the appellate court containing hallucinations. There were 12 fake case citations in the trial court filing. There were 9 in the filings on appeal. And, in my usage of fake case citations, I am including both situations where the case itself was fabricated into existence and situations where the case citation itself is correct, but the description of the contents is false. You can read the full order here:
Also in May, a lawyer was hit with non-monetary sanctions by a federal court in Maine for the problem of court filings with problematic case citations springing from use of either Claude or ChatGPT. (Fuller v. Hyde Sch., No. 2:25-cv-00354-SDN (D. Me. May 05, 2026). A Virginia lawyer had to file a declaration in federal court in San Francisco seeking to explain away his filing (using Claude Console) of a motion to quash a subpoena that contained false quotations attributed to real cases.
In June, 2 lawyers were suspended for six months from practice before the Ninth Circuit Court of Appeals and fined $2,500 for their persistent filings with the Court containing bad AI generated case citations. You can read that full order here:
And, in my own backyard in June, a plaintiff’s law firm was hit with monetary sanctions in federal court in Memphis in a lawsuit against another law firm that had previously represented it in a malpractice case. You can read the full opinion detailing all of the ways that the AI tool generated new content below and imposing monetary sanctions and a disciplinary referral here:
And then, also in June, there was a case out of Mississippi federal court where the lawyers on both sides of the litigation (4 lawyers in total) were sanctioned by the court (total monetary sanctions of $8,000). The two lawyers who actually used the AI tool had their pro hac vice admissions revoked. While the two local lawyers who did not fix any issues before filing were each suspended from practicing before the court for 2 years. All lawyers were also referred to disciplinary authorities. Because this one resulted in both sets of clients having to go get new counsel, it has been written about a good bit already lots of places. You can read that full opinion here:
Still just in June, the Michigan Court of Appeals entered a sanctions order against a lawyer with an extensive history in that case of filing things with the court with false content created through the use of GAI:
And, before June had wrapped up, the Third Circuit Court of Appeals lit into a Pittsburgh lawyer with a number of questions over problems created by his firm’s AI use in filings.