AI Overload in Courts: Surge in Filings, Fake Citations, and Legal Chaos

AI’s rapid integration into the legal system is causing unprecedented challenges, with both seasoned professionals and novices relying on tools like ChatGPT to craft persuasive arguments—sometimes with disastrous results. Last month, top law firm Sullivan Cromwell was forced to apologize after submitting a legal document riddled with fictitious case names, fabricated quotes, and incorrect citations to the U.S. Bankruptcy Code in a case involving an alleged scam operation based in Cambodia. The defendant denies the allegations.

This incident is not an isolated case. In 2025, a U.K. High Court case saw a barrister submit 18 fictitious case-law citations out of 45 total. Another 2025 disciplinary case involved a barrister using AI to prepare for a hearing while attempting to conceal fabricated citations. The 2023 Mata v. Avianca case remains one of the most high-profile examples, where an attorney used ChatGPT to draft a legal filing relying entirely on nonexistent judicial precedents.

New research is now quantifying AI’s impact on the legal system. A recent study highlights a significant rise in caseloads across U.S. federal courts. Anand Shah, a researcher at the Massachusetts Institute of Technology (MIT), led the study, which found that the share of pro se (self-represented) civil cases jumped from 11% to 18% in the post-AI era.

Shah and his co-author, Joshua Levy of the University of Southern California, analyzed a random sample of 1,600 filings from an eight-year period. Their findings revealed that AI-generated text in complaints surged from 0% before generative AI to approximately 18% in early 2026. “We were just floored,” Shah admitted.

Further analysis showed that the increase in AI-generated filings was concentrated in simpler, more templatable case types—such as debt collection or small claims—rather than complex areas like patents or securities law. Shah suggests this trend may indicate that AI is enabling individuals to pursue cases they previously wouldn’t have attempted, as generating legal arguments and documents has become far easier with minimal effort.

While anecdotal evidence points to AI straining the legal system, Shah notes that broader disruptions have not yet fully materialized in the data. “Cases are not resolving any faster or slower, which itself is a little surprising,” he says. However, he warns that the back-and-forth between opposing parties is increasing, dramatically expanding the number of filings judges must review—up by roughly 158%.