Morgan & Morgan Got Caught Submitting AI-Hallucinated Cases to a Federal Court
They built the AI tool themselves, let it invent fake citations, and signed the brief without a second look. Now two courts have had enough.
The case is Wilder v. President & Fellows of Harvard College, and the attorney in question is T. Michael Morgan, who was trying to get permission to appear in Massachusetts court despite not being licensed there. The judge said no. But it wasn’t just a simple denial. The ruling reads more like a detailed takedown of everything Morgan did wrong, and there’s a lot to work with.
It starts in Wyoming. In January 2025, Morgan was part of a legal team suing Walmart over injuries allegedly caused by a defective hoverboard toy. His associate drafted a set of pretrial motions using Morgan & Morgan’s own in-house AI platform, a tool the firm built and branded as “MX2.law.” The problem? MX2.law hallucinated eight of the nine cases cited in those motions. They were completely made up. Fake. Nonexistent. And every single one of them made it into the court filing.
When Walmart’s legal team flagged the bogus citations, Judge Kelly H. Rankin issued an order demanding that the attorneys explain themselves. The Morgan & Morgan lawyers withdrew the motions and admitted to the court that “the cases were not real and had been hallucinated by an AI platform.” That admission alone is staggering. This is America’s largest personal injury law firm, billing itself as the champion of everyday people, and it built an AI tool that fabricated case law and then let that fabricated case law go straight into a federal court filing.
What makes it worse is how it happened. Morgan’s associate drafted the motions using MX2.law, the AI generated fake citations and embedded them directly into the document, and then nobody, not Morgan, not his local counsel, reviewed a single word before signing off. They just affixed their e-signatures and sent it in. Judge Rankin found that the citations were so oddly formatted and truncated that a basic review would have raised obvious red flags. Morgan didn’t even look.
Three attorneys were ultimately sanctioned and fined a combined $5,000. Morgan’s associate Rudwin Ayala got hit the hardest, fined $3,000 and removed from the case entirely. Morgan himself was fined $1,000. The Wyoming court called his failure to review what he signed “a nondelegable duty” that he simply abandoned.
Now here’s where it gets even more embarrassing. When Morgan tried to get into the Massachusetts case, he disclosed the Wyoming sanctions in his application. Fine, points for honesty. But he said absolutely nothing about what he had done to fix the problem. No new review process. No changes to how MX2.law gets used. No acknowledgment that the firm had taken any steps to make sure fake AI-generated cases never end up in a court filing again. The Massachusetts judge called that omission “surprising” and “troubling,” which in judicial language is about as close to “you have some nerve” as it gets.
Then Morgan made things worse by botching the application itself. To appear in Massachusetts court as an out-of-state attorney, you need a Massachusetts-licensed lawyer to file the motion on your behalf. Morgan filed it himself, which means he was practicing law in a state where he has no license and no permission. He also paid the wrong fee, submitting $100 to the Board of Bar Overseers when the required amount is $355. The judge pointed out that a simple read of the rules would have made this obvious.
The court concluded that all of this together suggested Morgan had not learned from Wyoming and could not be trusted to follow Massachusetts law. Motion denied.
Morgan & Morgan has built its entire brand around being the firm that fights for the little guy. “For the People” is literally their slogan. But when your own attorney is getting sanctioned for filing AI-hallucinated case law he never read, then shows up in another state’s court, files the motion himself without a license, and pays the wrong fee, it raises a pretty uncomfortable question: how much of that brand is real, and how much of it is just a very good marketing budget?
