AI Workplace Lawsuits Flood the System Employers Built for People

Generative AI tools helped push the share of federal civil cases filed by people without lawyers from 11.33% before ChatGPT’s release to 16.94% afterward, according to the SSRN paper “Artificial Access to Justice.” In federal employment suits specifically, unrepresented plaintiffs filed 4,388 cases in 2025, up from 2,052 in 2021, against a record 26,635 total federal employment filings.

The corollary is that the courts, defense firms, and judges are absorbing a flood of filings the system was never built to handle at machine speed. Employers still win almost every one of these cases. The cost has moved elsewhere: into motion practice, into discovery disputes, and into the slow work of verifying every citation a chatbot invents.

The Filings Arrived Faster Than the Courts

The headline number comes from Lex Machina’s 2026 Employment Litigation Report. Pro se plaintiffs went from 9.7% of federal employment suits in 2021 to 16.5% in 2025, more than doubling in raw count from 2,052 to 4,388 cases over a window when total employment filings set a multiyear record. Those figures were corroborated by an empirical analysis of roughly 2.8 million federal civil filings, which found the federal civil pro se plaintiff rate rose from 11.33% pre-GenAI to 16.94% post-GenAI.

Composition matters as much as counts. The empirical study found AI-flagged complaints are disproportionately filed by first-time rather than repeat litigants, more citation-dense than human-drafted predecessors, and geographically unevenly distributed across federal districts. The pool of people who can now walk into a federal courthouse with a polished complaint is bigger than the legal profession expected.

• 49% jump in pro se employment lawsuits last year (Fisher Phillips via Bloomberg Law).
• 16.5% of all federal employment suits in 2025 were pro se, up from 9.7% in 2021.
• Pro se plaintiffs lose on the merits at a ratio greater than 40 to 1 (Lex Machina).
• Only 29% of pro se cases settle, compared with 77% of represented cases.

Workers Were Already Armed; Chatbots Just Made the Trigger Cheap

The surge is not just a supply story. It is also a tone story. Detroit employment attorney Brian D. Shekell told Michigan Lawyers Weekly’s read on the Lex Machina report that AI lets workers draft responses that used to give themselves away. “Before, you would be able to tell pretty quickly if an employee had an attorney that was representing them,” Shekell said. “Now it’s a little bit more difficult to discern because with Chat GPT and other AI platforms, employees are able to draft responses or write letters to the employer that uses a lot of legal … jargon and buzzwords.”

The same attorneys describe a parallel problem on the demand side. Royal Oak employment attorney Jennifer L. McManus flagged what she called the “pleasing” tendency of large language models, in which a chatbot reads the user’s own framing of a dispute and tells them what they want to hear. Inflated settlement asks of hundreds of thousands of dollars are increasingly common in pro se matters, because the model has validated the user’s worst-case number.

Workers are also counter-punching from a position of long-running surveillance. The same AI revolution that lets staff file grievances from a kitchen counter has, for years, scored their keystrokes, ranked their call-handling times, and decided who gets an interview. According to the National Employment Law Project’s bossware policy agenda, webcam monitoring software uses AI to flag workplace violations in real time. The lawsuit is now a written response, with the same tooling, to a decade of algorithmic management.

The Defense Bill the System Can’t Avoid

Employers still win, on paper, at 40 to 1. The bill is paid in time. Bloomberg Law’s coverage of the pro se cost surge quotes Kristin White, a Denver-based Fisher Phillips partner: “Every litigator in my office is handling at least one case brought by a pro se plaintiff.” Her firm estimates these defenses run about 10% to 15% more than a typical employment claim, because the litigants make larger settlement demands, file more motions, and pursue longer discovery.

That number balloons when the filings include fabricated citations and duplicative motions. A judge in the Central District of California ordered a Chinese plaintiff to pay Arnold & Porter more than $66,000 in attorney fees after the firm spent time chasing fake cases and responding to bad-faith filings. The plaintiff had told the court he was training five more Chinese nationals to use AI to sue the same defendants. The sanction did not interrupt the mechanism. The litigant appealed, and his opening brief in the appeals court ran 456 pages, most of them motions already rejected below.

Sanctions Are Coming, Slowly

Courts are starting to push back, but in patches. According to the AI hallucination cases database maintained by Damien Charlotin at HEC Paris, more than two dozen pro se litigants have been hit with monetary sanctions since mid-2023, and more than half of those fines were levied since December 2025. A Bloomberg Law tally recorded 52 court rulings in February alone identifying improper AI use, up from 2 a year earlier.

The most aggressive remedies are coming in the form of standing orders and relief from response obligations, per the Baker Donelson analysis of AI-assisted pro se litigation. Employers now have a usable playbook, and judges are using it.

• Allen v. Casper (N.D. Ill., March 2026): $1,500 Rule 11 sanction after a 112-page opposition brief cited entirely fictitious cases.
• Tantaros v. Fox News Network (S.D.N.Y., March 2026): filing struck, warning issued that future AI abuse will trigger sanctions.
• Thomas v. Delaware Technical and Community College (D. Del., Nov. 2025): employer relieved of any obligation to respond after more than 49 filings.
• Rako v. VMware (N.D. Cal., Feb. 2026): standing order barring filings with AI-hallucinated citations.

Some legislatures are taking a different angle. New York state legislators are weighing a bill that would ban generative AI tools from providing legal advice outright and allow civil suits against chatbot owners who violate the law. A separate, more invasive question is working its way through district courts: whether a litigant’s private chats with ChatGPT or Claude are discoverable. Recent rulings in United States v. Heppner, Warner v. Gilbarco, and In re OpenAI, Inc. Copyright Infringement Litigation are sketching the boundaries of that doctrine.

The Paradox at the Center

The deeper finding is that AI-assisted complaints look better and fail faster. The SSRN paper “Artificial Access to Justice” found that AI-flagged complaints are more likely to be dismissed and to terminate at earlier procedural phases than human-drafted counterparts. That holds even though the filing rate has more than doubled in four years.

The thing about AI is that it aims to please. So, when people are feeding their own scenarios into AI, first of all, they’re feeding their scenarios in and shining the best light on themselves and what they believe their claims to be. And then AI is spitting out something that it thinks will please you.

Jennifer L. McManus is a Royal Oak, Mich., employment attorney quoted in Michigan Lawyers Weekly. The bottleneck in the system, in other words, is not outcomes. It is endurance: AI is keeping weak cases alive long enough to extract settlement value from employers tired of motion practice, and judges are now the only ones with a fast enough tool to push back.

The Same AI Is Now on the Other Side of the Filing Window

The legal system is also judging AI when it is the employer’s tool. Roughly 99% of Fortune 500 companies now use AI to filter job applicants, and roughly 40% of companies expect to use AI to conduct screening interviews of candidates, per Bricker’s 2026 AI hiring litigation roundup. The Mobley v. Workday litigation alleges the platform’s screening tools incorporated data points like medical leave history and gaps in employment that correlate with disability and age, in violation of the ADEA, the ADAA, and Title VII. The federal court in Mobley allowed those claims to move forward at the start of 2026, rejecting Workday’s motion to dismiss.

A separate class action filed January 20, 2026, against Eightfold AI reframes the same dispute as a consumer protection problem. Two applicants allege the platform generated secret “likelihood of success” scores on a 0 to 5 scale and dossiers that functioned as undisclosed consumer reports, in violation of the FCRA and California’s Investigative Consumer Reporting Agencies Act.

Lower-stakes cases are already producing outcomes. The U.S. Department of Justice imposed a nearly $10,000 fine on an IT company for posting AI-generated job advertisements that unlawfully excluded U.S. citizens. The legal system is now ruling on AI-produced content from both sides of the docket: AI-drafted complaints from workers, and AI-generated hiring decisions from employers.

What the Courts and States Are Building to Catch Up

State legislatures have moved into the gap left by federal regulators. More than 20 new California AI laws took effect January 1, 2026. Connecticut Governor Ned Lamont signed broad AI legislation on May 27, 2026, requiring employers to disclose when automated employment-related decision tools interact with applicants or employees; disclosure rules become enforceable on October 1, 2027. Colorado Governor Jared Polis signed SB 26-189 on May 14, 2026, repealing and reenacting the Colorado Artificial Intelligence Act and substantially altering obligations for employers using AI in employment decisions; the new framework takes effect January 1, 2027.

California Governor Gavin Newsom issued an executive order targeting AI-driven labor market disruption and directing state agencies to recommend updates to the California WARN Act for AI-related mass layoffs. The order recognizes a problem the rest of the country is only starting to see: the AI in question is also filing paper, in bulk, on the other side.

Federal policy is moving in the opposite direction. A December 11, 2025, executive order signed by President Donald J. Trump, “Ensuring a National Policy Framework For Artificial Intelligence,” attempts to restrict states from independently regulating AI in ways the order calls “onerous and excessive.” If the patchwork keeps growing while federal preemption claims keep moving, employers face compliance uncertainty and courts face a litigation architecture that was not built with AI as a party on either side.

Frequently Asked Questions

How many workplace lawsuits are now filed by people without lawyers?

Pro se plaintiffs filed 4,388 federal employment lawsuits in 2025, up from 2,052 in 2021, per Lex Machina’s 2026 Employment Litigation Report cited by Baker Donelson. That lifted their share of all federal employment suits from 9.7% to 16.5% over four years.

Are workers winning more when they use AI to file?

No. Lex Machina reports pro se plaintiffs still lose on the merits by a ratio greater than 40 to 1, and the SSRN paper “Artificial Access to Justice” found AI-flagged complaints are more likely to be dismissed and to terminate at earlier procedural phases than non-AI-flagged filings.

Why are these lawsuits more expensive to defend if employers usually win?

Defense costs rise because AI-assisted pro se litigants make larger settlement demands, file more motions, and pursue longer discovery. Fisher Phillips partners put the defense cost uplift at 10% to 15% over a typical employment claim.

Are courts sanctioning people for AI misuse in filings?

Yes. More than 24 pro se litigants have been hit with monetary sanctions since mid-2023, with more than half of those fines levied since December 2025, per the AI hallucination cases database maintained by Damien Charlotin at HEC Paris. A Bloomberg Law tally put the count of court rulings identifying improper AI use at 52 in February alone.

Could AI legal advice be banned outright?

New York state legislators are weighing a bill that would ban generative AI tools from providing legal advice and allow civil suits against chatbot owners who violate the law. No state has finalized such a ban as of mid-2026.