Since the Ninth Circuit suspended two attorneys in June and California's State Bar charged three more in May for AI-fabricated citations, the legal profession has focused on licensed lawyers getting caught. But the problem extends far beyond that.

The larger issue is coming from people who never had a lawyer to begin with.

The Numbers Are Real

Researchers Anand Shah at MIT and Joshua Levy at the University of Southern California studied 4.5 million federal civil cases filed between 2005 and 2026.

Self-represented litigants — people suing without an attorney — climbed from 11% of federal civil cases in 2022 to 16.8% in 2025. The raw number of filings from those individuals more than doubled compared to pre-2023 levels.

The AI fingerprints are visible throughout. The researchers ran 1,600 randomly sampled court documents through Pangram, a commercial AI-text detector. The share flagged as containing AI-generated writing jumped from 1% in 2023 to 18% in 2026.

What a Federal Judge Actually Sees

Judge Maritza Braswell, a federal magistrate in Colorado, deals with these filings daily. Her perspective cuts through the noise on both sides.

"I do correlate that to AI," Braswell told MIT Technology Review, "in part because I see AI use." She flags hallucinated case citations and fabricated quotes. But she also notes something the sanction-focused crowd often overlooks: some of these AI-assisted filings are better than what she used to get.

"I'm also actually seeing better-drafted pleadings," she said. Handwritten, barely-coherent pro se filings used to consume enormous amounts of judicial time. Now judges can at least understand what people are arguing — even if the underlying case is weak.

She processes AI-drafted motions faster. That represents a real efficiency gain in an overwhelmed court system.

The Distinction Being Overlooked

Mainstream coverage conflates two separate issues: the sanctions story and the access-to-justice story. They are not the same.

When a paid attorney submits AI-fabricated citations, that's malpractice. Those lawyers took money from clients, had a professional duty, and failed it. The Ninth Circuit was right to suspend them.

When an unemployed person with a $3,000 wage theft claim and no money for a lawyer uses ChatGPT to draft a complaint — that is a completely different situation. The same AI risks exist (hallucination, fake citations). But the moral and legal calculus differs substantially.

Judge Braswell flagged this distinction directly. Courts are already required to read pro se filings charitably. That obligation does not disappear because AI wrote the draft.

What Courts Are Actually Doing

Federal courts have been rolling out local rules requiring attorneys to disclose AI use in filings — a trend the ABA Journal has tracked across multiple districts. Those rules target licensed lawyers. They do NOT meaningfully address self-represented litigants who have no bar license to protect and no ethical obligations beyond not committing fraud.

The gap is widening. Attorneys face escalating sanctions for AI misuse. Ordinary citizens face nothing except having their cases dismissed for bad citations. Two different accountability systems running in parallel.

The Real Risks — On Both Ends

AI is NOT improving self-represented litigants' odds of winning. The MIT/USC study found that despite better-drafted pleadings, success rates for pro se plaintiffs haven't meaningfully improved. People are articulating losing cases more clearly. That's cleaner paperwork, not justice.

There's also a clog risk. More filings, even if individually better-drafted, means more judicial workload. Federal courts are already backlogged. A doubling of pro se filings in three years could strain the system regardless of filing quality.

Hallucinations remain a real problem. Braswell said she has to check these filings carefully, and that verification work takes time. The efficiency gains from cleaner prose can be offset by citation-checking.

What This Actually Means

Lawmakers are starting to ask who's liable when an AI chatbot gives someone bad legal advice that tanks their case. That's a legitimate question — and one that won't be answered cleanly by bar association ethics rules written for licensed attorneys.

The courts, Congress, and state legislatures are at least two years behind this problem.

For regular people: if you're using AI to help file a legal claim, understand that the chatbot has no duty to give you accurate advice. It will sound confident while being completely wrong. Verify every case citation independently before it goes anywhere near a court. Judges are catching these errors. And unlike an attorney, you have no one else to blame.

The technology opened a door that was genuinely closed to millions of people who couldn't afford legal help. But a door into a room full of land mines remains a dangerous door.