A recent study indicates a significant rise in the use of AI-generated text in legal filings by self-representing plaintiffs, known as pro se litigants. Conducted by Anand Shah from MIT and Joshua Levy from USC, the research highlights that 18% of pro se filings now include text produced by large language models (LLMs). This increase coincides with a 64% surge in pro se docket entries within the first 180 days of litigation since the introduction of widely accessible LLMs.
Historically, most pro se filings have been submitted by incarcerated individuals, but the study reveals a notable increase in filings from non-prisoners, climbing from 11% to 16.8% by fiscal year 2025, a record not seen in the past 25 years. A Minnesota federal judge, Patrick J. Schiltz, has labeled this trend as an “existential threat to the federal courts,” citing concerns over the volume of frivolous lawsuits that burden the judicial system.
From 1998 to 2017, pro se plaintiffs faced a staggering 96% loss rate in their cases, raising questions about the implications of AI in legal contexts. The New York Times has drawn attention to these challenges, illustrating the potential misuse of AI in generating unmeritorious lawsuits that could overwhelm court resources.