Federal judges across the United States are experiencing something unprecedented: a measurable surge in self-represented civil lawsuits, and growing evidence that large language models are fueling it. A new study examining 4.5 million federal civil cases filed between 2005 and 2026, published by researchers at MIT and the University of Southern California, found that the share of suits brought by people without legal counsel increased from around 12 percent to nearly 20 percent over the study period — a shift that correlates with the widespread availability of AI chatbots.
Judge Maritza Braswell, a federal magistrate in Colorado whose work was featured in MIT Technology Review, says the trend is unmistakable in her chambers. She reviews stacks of filings from self-represented litigants daily and has trained herself to recognize the stylistic fingerprints of AI-generated prose. "I do correlate that to AI in part because I see AI use," she told the publication. The tells are familiar to anyone who has read LLM output carefully: a certain structural formality, overuse of parallel constructions, and occasionally — more damagingly — citations to cases that do not exist.
Expanded Access, Uncertain Outcomes
The picture emerging from the research is nuanced. AI tools are unambiguously lowering the barrier to participating in the legal system. People who previously could not afford an attorney, or whose claims were too small to attract one, can now draft coherent pleadings and navigate procedural requirements that once seemed impenetrable. In that sense, the technology is expanding access to justice in a meaningful way.
However, expanded access does not appear to be translating into improved win rates for the people filing these suits. Judges report that while AI-assisted briefs are often better structured than handwritten equivalents, the underlying legal reasoning can be flawed in ways that are harder to spot at a glance — including the hallucinated case citations that have already generated formal sanctions in several high-profile instances. The legal system is not designed to flag every incorrect citation as it comes in; discovery happens after filings have already consumed court resources.
Questions of Accountability
The broader implications extend beyond workload management for federal clerks. Courts are beginning to wrestle with a genuinely novel question: what legal status, rights, or responsibilities should LLMs bear when they function as de facto legal counsel? If an AI model provides incorrect legal advice that leads a litigant to take a damaging procedural step, who is accountable?
These questions remain largely unresolved, and they are arriving faster than the legal profession was expecting. Several courts have issued standing orders requiring litigants to disclose AI assistance in filings; some judges have gone further and required certification that AI-generated citations have been independently verified. But consistent national standards do not yet exist.
For enterprise technology and legal teams, the trend carries concrete implications. Organizations deploying AI tools that could be used in legal contexts — contract review, compliance documentation, dispute correspondence — face increasing pressure to build verification and disclosure frameworks proactively, ahead of regulation rather than in response to it.
Why It Matters
AI is quietly reshaping who can access the legal system — but it is also introducing new failure modes at scale. As self-represented AI-assisted filings increase, courts, enterprises, and technology developers will need to develop shared standards for accountability before the gap between capability and governance becomes a systemic problem.