AI-Drafted Lawsuits Reshape the Legal Career Ladder

The share of federal court documents containing AI-generated text jumped from 1 percent in 2023 to 18 percent in 2026, according to a study of 4.5 million federal civil cases reported June 4 by MIT Technology Review. The research, conducted by Anand Shah of MIT and Joshua Levy of USC, also found that self-represented filings climbed from 11 percent of cases in 2022 to 16.8 percent in 2025.

The surge is concentrated, not gradual. In Vermont, pro se filings rose from roughly 45 a year before 2022 to 1,100 in 2024. The researchers identified AI-drafted passages using Pangram, a text-detection tool built to flag content produced by models like ChatGPT and Gemini.

Judges describe a mixed picture. Maritza Braswell, a federal magistrate judge in Colorado, told MIT Technology Review she can "generally understand what they're arguing better with AI assistance," even as filings still "contain hallucinations and errors." Despite cleaner pleadings, win rates for self-represented litigants have not improved, and fabricated case citations remain a persistent problem.

The courts themselves are unsettled on the rules. Judges and legislatures are split on whether conversations with a chatbot qualify for attorney-client privilege or work-product protection, and on whether an AI tool that drafts a legal filing is engaged in the unlicensed practice of law. Nippon Life Insurance Co. sued OpenAI in March on that theory; OpenAI's motion to dismiss, filed in May, is still pending. New York lawmakers introduced a bill the same month that would bar chatbots from impersonating licensed attorneys.

What this means for job seekers

This is a clear signal about where legal work is heading, and it rhymes with the broader shift toward AI in the job search. When a litigant with no lawyer can produce a coherent complaint, the routine drafting that once anchored entry-level legal work — document prep, boilerplate filings, first-pass research — loses its scarcity value. That pressure lands hardest on paralegals and junior associates whose day-to-day is exactly the kind of repeatable output a model now generates in seconds.

But the study also points to where the durable jobs are. AI produces clearer arguments, yet it does not win more cases, and it keeps inventing citations that humans must catch. That gap is the opportunity. The skills courts still reward — verifying that a cited case actually exists, spotting hallucinated authority, building strategy for a contested matter, and managing the ethics questions judges are now grappling with — are precisely the ones AI cannot supply. Aspiring legal and paralegal workers should position around AI oversight, citation verification, and complex litigation rather than routine drafting. If you are weighing the field from scratch, treat this as a career-path decision about judgment work versus production work; the production layer is the part being automated.

Sources

• "How courts are coping with a flood of AI-generated lawsuits," MIT Technology Review, technologyreview.com — accessed June 5, 2026

• "Pangram AI content detection," Pangram, pangram.com — accessed June 5, 2026