A federal court in Vermont used to receive about 45 pro se lawsuits a year. In fiscal 2024, that number jumped to 1,100.
More than twentyfold.
On May 26, researchers from MIT and the University of Southern California revealed the cause. After analyzing 4.5 million civil cases (fiscal 2005–2026) and 46 million PACER court records, they concluded that ChatGPT is turning U.S. courts into an unstoppable lawsuit printer.
By the Numbers
First, the share of pro se cases (self-represented litigants):
- Past 20 years: stable at ~11%
- Fiscal 2025: 16.8%
In fiscal 2025 alone, there were 41,490 pro se cases—roughly double the pre-AI annual average. More critically, 59% of the recent increase in civil litigation comes from these self-represented plaintiffs.
How to prove AI wrote them? The researchers used an AI text detector called Pangram, first validating it on 1,600 random complaints, then running it on the full dataset. The detected AI footprint grew as follows:
- 2023: 1.0%
- 2024: 3.5%
- 2025: 10.5%
- Early 2026: 18.0%
That means roughly one in five lawsuits is now AI-generated.
A Well-Intentioned Start
To be clear, AI helping ordinary people with legal matters started with good intentions. Those who couldn't afford a lawyer often gave up facing a pile of incomprehensible legal procedures. Now you can ask ChatGPT to explain what's needed to file a case, organize facts, summarize precedents, draft motions, and generate formatted documents—almost for free. It fills the famous 'justice gap' in legal services.
The problem is the volume.
When the Barrier Falls, the Flood Comes
When filing becomes too easy, courts get clogged. The study found that docket activity in the first 180 days after a pro se plaintiff files has surged 158% compared to pre-AI levels. A barrage of weak, repetitive AI motions forces opposing lawyers to respond one by one, driving up client bills.
Federal Judge Patrick Schiltz of Minnesota put it bluntly:
'An existential threat to the federal courts.'
MIT researcher Shah was more direct, saying that if this trend continues, courts would 'basically have to shut down.'
Unintended Consequences
Researchers pinpoint the turning point as November 2022, when ChatGPT launched. The timeline matches—the subsequent spate of fabricated cases and AI-hallucinated citations making headlines also started then.
Courts can currently only add disclosure requirements and set up firewalls. But the fundamental contradiction remains: since AI has lowered the barrier to 'writing a plausible complaint' to the floor, it can't be blocked.
A tool meant to bridge the justice gap has instead clogged the courts themselves. U.S. judges have yet to find a way out.
Sources: CocoLoop, The AI justice gap solution is slowly turning into an existential paperwork nightmare for US federal courts (The Decoder); People without lawyers are using AI to flood courts with lawsuits (TechSpot)