In a nutshell: It's a case of another day, another problem caused by the proliferation of generative AI. More and more people are using the technology to file lawsuits in which they represent themselves, and, unsurprisingly, courts are being clogged with many of these frivolous suits.
The New York Times highlights the case of Donald Sauve, a Minnesota man who sued his ex-wife, her lawyer, and a state judge after an earlier legal challenge was rejected as frivolous. His first handwritten complaint, which sought $275,000 in damages, was dismissed in less than a month for lack of jurisdiction.
Three months later, Sauve returned with help from ChatGPT and Claude. This time, his complaint was neatly typed and accompanied by 50 additional filings, including what he called a "case law synthesis" supporting his claim. The case was dismissed, again, but before that happened, every filing had to be read, captioned by a clerk, and entered into the public docket.
Sauve's example is a perfect illustration of the problem courts are now facing. AI has made it easier for people without lawyers to produce documents that look and sound like they were prepared by legal experts, even when the underlying claims are weak, confused, or baseless. In Sauve's case, Judge Patrick J. Schiltz, chief of Minnesota's US District Court, ordered that any further filings would be "shredded without any additional notice."
Lawsuits filed by people representing themselves rather than using a lawyer are called pro se cases. They are common in federal courts, but most come from prisoners using prison law libraries to challenge conditions, convictions, or alleged civil rights violations. The AI-driven concern focuses more on non-prisoner pro se litigants. These are usually ordinary people who either can't afford legal help or believe they can argue their own case, and can now use chatbots to generate large volumes of court-ready paperwork.
A new working paper from MIT's Anand Shah and USC's Joshua Levy examined the scale of the issue. Looking at more than 4.5 million non-prisoner federal civil cases from fiscal 2005 through fiscal 2026, along with 46 million PACER docket entries, the researchers found that non-prisoner pro se cases rose from a long-term average of around 11% to 16.8% in fiscal 2025.
The study also found that pro se cases are creating more work once they enter the system. The volume of docket entries per court generated by these cases in the first 180 days had risen 158% above the pre-AI average by 2025. In a sample of 1,600 complaints from 2019 to 2026, more than 18% of 2026 complaints were flagged as likely containing AI-generated text.
The counterargument here is that AI could help people who can't afford attorneys access a legal system that is often intimidating and expensive. Judge Michael Y. Scudder of the Seventh US Circuit Court of Appeals wrote this year that the technology offers "great promise" for improving access to justice.
But courts are already warning that self-represented litigants remain responsible for AI-generated errors. In January, the Seventh Circuit said accuracy and honesty still matter after a pro se filing appeared to contain AI-generated false citations.
As we reported last year, AI-generated legal filings were already making a mess of the judicial system by producing fake cases, phantom precedents, and fabricated citations. Today, the problem has moved beyond lawyers cutting corners to almost anyone with access to a chatbot.