Patent infringement claims have long been considered some of the most difficult lawsuits to win. They tend to be lengthy, demanding copious legal research and complex technological analysis. While it’s always nice to harness the power of technology to help dig through case law and other research, it’s also important to double-check the results to ensure that AI-generated citations don’t manage to hamstring your case and get you in trouble.
Yes, it’s yet another edition of “Lawyers Get Tripped Up by Filing a Brief Containing AI Hallucinations,” as trust in AI tools for legal citations continues to be perhaps a little higher than it should be. That might well be the revised opinion of five attorneys working as co-counsel for plaintiffs Lexos Media IP LLC in its patent infringement lawsuit against Overstock.com. The suit, currently in pre-trial motions in the District of Kansas, was forced to make a slight detour in December 2025 after Federal Judge Julie Robinson issued an order to show cause for using generative AI in their response to a motion to exclude testimony given by one of their expert witnesses.
Her ruling on February 2, 2026, contained bad news for those involved. While the fines, public admonishment through the court order, and the requirement that one of the attorneys self-report to his state bar for possible further disciplinary action for violations of Federal Rule of Civil Procedure 11 are bad enough, not being permitted to submit a revised reply brief is perhaps even more devastating.
Wait, I Thought You Were Going To Check It (x5)
As just about anyone who has practiced law can tell you, sifting through precedent in support of an argument is necessary but often tedious. The use of generative AI tools to aid legal professionals in finding case citations was understandably met with great hope and excitement. So far, the results have been mixed.
While AI tools have shown promise in the legal field as search engines, draft creators, and in other unexpected ways, they have also returned results containing misstatements, hallucinations, and references to nonexistent cases. While many are likely caught by legal teams before being submitted to the court, enough have gotten through for judges to lose patience and punish the litigants who slip up.
The patent infringement suit filed by Lexos Media focuses on technology for modifying the display of a cursor image on websites that Overstock.com allegedly used without permission. On June 13, 2025, Overstock.com filed two motions to exclude testimony given by Dr. Samuel Russ, Ph.D., an expert witness for the plaintiffs. Questioning apparent mistakes made by Russ concerning the definition of a cursor, Overstock.com also filed a motion for summary judgment under Daubert on the grounds that, without Russ’s testimony, Lexos Media couldn’t prove infringement.
Lexos Media’s legal team filed a response on July 7 to challenge the exclusion, offering several citations in support of its position. Overstock.com’s attorneys returned fire by alerting the court to multiple instances of nonexistent quotations, misrepresentations about cited authority, and nonexistent and incorrect citations in Lexos Media’s filing. The cases referenced, including those involving Microsoft and Walmart, were real. However, the citations provided were either incorrect or nonexistent.
In the order to show cause, District Judge Robinson demanded that counsel explain how something like this could have happened. It turns out that nobody got around to checking the citations.
Sanctions Add Up Quickly
Judge Robinson ruled that five attorneys were in violation of Fed. R. Civ. P. Rule 11(b) for not making a reasonable inquiry into the results presented by their AI tools before signing and submitting their response to the court. The attorney given the harshest sanctions was the author of the troubled filing. The lawyer admitted he didn’t check the AI-generated results, apologizing while citing extenuating family issues and his fear of missing a deadline as rationale. He was fined $5,000, had his pro hac vice admission for the case revoked, and was ordered to self-report the ruling to his state bar. He must also submit documents showing how his law firm will institute new procedures to ensure it doesn’t happen again.
Three attorneys each received $3,000 fines for signing off on false content. Only the junior attorney who conducted a technical edit of the filing was not fined.
The final attorney voluntarily self-sanctioned himself by promising not to serve as local counsel or a legal authority for pro hac vice attorneys for a 12-month period. His firm agreed to upgrade its AI-review policies.
The Worst Penalty Last
The consequences of their actions (or inactions) may strike a much more serious blow to the plaintiff’s lawsuit. Judge Robinson denied the request to amend the reply. Instead, the fictitious citations created by the generative artificial intelligence research tools will be struck and not replaced, perhaps moving Overstock.com one step closer to receiving its requested dismissal by summary judgment.
The ruling may have a ripple effect on the legal world, prompting firms to pay closer attention to citations generated by AI searches.
Related Resources
- Haven’t Started Using AI Yet? You Aren’t as Behind as You Might Think. (FindLaw’s Practice of Law)
- Patent Infringement and Litigation (FindLaw’s Intellectual Property Law)
- Alabama’s Million-Dollar Prison Defense Team Caught Using Fake Citations (FindLaw’s Practice of Law)
