The Chatbot in the Courtroom

On a single day in February, two federal courts in the United States handed down completely contradictory rulings on a question that didn't exist a few years ago: Are your private conversations with an AI chatbot protected by the law?

A court in Michigan said yes, ruling that a self-represented litigant’s chats with ChatGPT were shielded "work product." Meanwhile, a New York court said no, arguing that since AI companies can share user data, there is no reasonable expectation of confidentiality. This judicial split highlights a quiet revolution happening in courtrooms: the explosion of the AI-assisted lawsuit.

For citizens who cannot afford legal representation, large language models have become digital co-counsels. A joint study by MIT and USC researchers analyzed 4.5 million federal civil cases, revealing that self-represented lawsuits jumped from 11% of the docket in 2022 to 16.8% by 2025. By 2026, the share of these court documents flagged for AI-generated text skyrocketed to 18%, up from just 1% in 2023. The impact is localized but intense. In Vermont, for instance, self-represented cases surged from around 45 annually to over 1,100 in 2024, driven largely by a viral Reddit guide teaching immigrants how to use Microsoft Copilot to sue over delayed applications.

For judges, this AI wave is a mixed blessing. Colorado federal magistrate judge Maritza Braswell notes that while she has to watch out for AI hallucinations and fabricated quotes, the technology actually helps her do her job. Pro se filings used to be notoriously cryptic; now, AI helps litigants articulate their arguments clearly, allowing judges to better understand and assist them.

However, sounding like a lawyer isn't the same as being one. As California judge Allison Goddard discovered when a slip-and-fall plaintiff demanded an absurd $700,000 settlement based on ChatGPT's advice, AI often dispenses terrible legal counsel. Furthermore, the MIT/USC study revealed a sobering truth: despite having better-drafted pleadings, self-represented individuals are not winning their cases at a higher rate.

AI has undeniably lowered the barrier to entry for the justice system, translating the complex language of the law into something accessible for the average person. But as courts struggle to decide whether a chatbot is a legal confidant or a liability, one thing remains clear: navigating the legal system requires strategic, multifaceted expertise that no language model can currently generate.

Key Points

• AI-assisted legal filings by self-represented individuals have surged dramatically in recent years.
• Federal courts are currently divided on whether conversations with AI chatbots are protected under legal privilege.
• Judges find AI-drafted documents easier to read, though they must remain vigilant for AI hallucinations.
• Despite clearer arguments, AI assistance does not improve the actual win rates for people without lawyers.

Why It Matters

The influx of AI in the legal system is exposing critical gaps in legal ethics, forcing society to decide what rights and responsibilities algorithms hold when they act as pseudo-lawyers.

Sources:

• How courts are coping with a flood of AI-generated lawsuits — MIT Technology Review - AI