AI Invades American Courts: A Challenge for Justice

The Rise of Pro Se Filings

In the chambers of Judge Maritza Braswell, a federal magistrate in Colorado, stacks of documents are piling up. These documents are filed by individuals representing themselves, often because they cannot afford to hire a lawyer or because their cases are deemed too weak to attract a professional's attention. Judge Braswell, like many of her colleagues across the United States, has noticed a significant increase in these filings. A recent study analyzed 4.5 million federal civil cases from 2005 to 2026, revealing that the share of lawsuits filed by pro se litigants rose from 11% in 2022 to 16.8% in 2025. The number of filings has more than doubled compared to pre-2023 levels.

Judge Braswell attributes this increase to the use of artificial intelligence (AI). She explains, “I connect it to AI partly because I see its use.” As a judge who uses AI to review court documents, she has learned to recognize the writing style of large language models. She can identify instances where hallucinations or fabricated citations appear. She also notes that some filings are better written thanks to AI.

However, while AI seems to broaden access to justice, it does not appear to improve the success rates of plaintiffs. Judges are beginning to question the rights and responsibilities that large language models should have when they replace lawyers. For instance, they wonder if a chatbot has a duty to provide sound advice, as a human lawyer would. An increasing number of lawmakers in the United States are questioning who should be held accountable when chatbots give poor legal advice.

AI and the Surge in Lawsuits

To test whether AI was behind the increase in lawsuits filed by pro se litigants, the study's authors, Anand Shah from MIT and Joshua Levy from the University of Southern California, analyzed 1,600 randomly sampled court documents using Pangram, a commercial AI text detector. The share of documents flagged as containing AI-generated writing rose from 1% in 2023 to 18% in 2026.

For Judge Braswell, this is not necessarily a cause for concern. While the increase in AI-assisted filings may add to their workload, she and many other judges find that cases are easier to resolve because AI helps untrained individuals articulate their arguments more effectively. Court documents filed by pro se litigants are notoriously difficult to decipher. Some arrive in the form of handwritten scrawls bordering on gibberish that judges take time to decode. Regardless of their complexity, judges are required to read them with understanding.

These days, Judge Braswell processes AI-written motions faster than those penned by litigants. “I have to be really careful because some of them contain hallucinations and errors, but I can generally understand what they are arguing better with the assistance of AI than without it,” she says. Clearer filings allow Judge Braswell to hear them better. “If I understand an argument a little better, I will probably be able to help a little more,” she adds.

Online communities are emerging to share self-help guides on using AI to file lawsuits. In December 2024, a viral post on Reddit guided asylum seekers to sue the U.S. Citizenship and Immigration Services for delays in processing their applications: draft a mandamus petition using Microsoft Copilot, pay a lawyer $150 to improve it, and file in the District of Vermont. The number of pro se filings in Vermont jumped from about 45 per year before 2022 to over 1,100 in 2024.

Despite this, pro se litigants are much more likely to lose their cases than those who have legal representation, and this does not change even with the addition of AI, according to the study. “It turns out that filing a lawsuit is a complex and multifaceted task. It’s not just about writing text,” explains Levy.

The Client-Chatbot Privilege Question

Judge William Garfinkel, a federal magistrate in Connecticut, has been on the bench for three decades, pondering all sorts of questions about the relationship between lawyers and their clients. Recently, he has wondered whether conversations between individuals and chatbots providing legal advice should be protected, like those with lawyers. “One could make a good argument that... conversations with large language models like Claude or ChatGPT should receive some protection,” he says.

Courts are beginning to consider this question. In February, a federal court in Michigan ruled that a pro se litigant's conversations with ChatGPT to prepare their case were considered work product — a legal work protected from opposing parties. The decision came on the same day that a federal court in New York ruled that documents generated by a criminal defendant using Claude were not privileged communications between attorney and client nor work product. The court maintained that Claude is not a lawyer and that a user does not have a “reasonable expectation of privacy in their communication” with it, as AI companies may disclose user data to third parties.

In March, Judge Braswell ruled that the use of a chatbot by a pro se litigant should remain off-limits. “It is true that AI systems like ChatGPT, Claude, Gemini, and others... collect user data for training and other purposes. But... that does not eliminate all expectations of privacy,” she wrote. Courts have since remained divided on the issue.

The Responsibility of Chatbots

Some judges are also questioning whether a chatbot, like a lawyer, has a duty to provide sound legal advice. Judge Allison Goddard, a federal magistrate in California, has noticed that pro se litigants often receive poor advice from ChatGPT when trying to assess the value of their case during settlement negotiations. In one instance, a plaintiff who slipped and fell in a store demanded $700,000 from the store, which was far above the case's value.

“Where do you get the idea that you’re going to get $700,000? Did you go to ChatGPT?” Judge Goddard asked. “Well...” the plaintiff stammered. She then explained the law to the individual to show why ChatGPT was wrong and suggested a lower amount. “It’s like Dr. Google went to law school,” she says.

Then there’s the question of who is responsible when a chatbot makes such errors. In March, Nippon Life Insurance Company sued OpenAI, alleging that ChatGPT practiced law without a license and helped a woman reopen a previously settled lawsuit, flooding the court with frivolous filings. “ChatGPT is not a lawyer,” the lawsuit stated.

In May, OpenAI moved to dismiss the case, arguing that ChatGPT does not practice law. “ChatGPT is not a person and does not possess or use any degree of legal knowledge or skill,” OpenAI stated in its filing. The case is still pending in court.

States have begun considering legislation that would hold AI companies accountable when their chatbots provide poor legal advice. New York introduced a bill in March that would prohibit chatbots from impersonating lawyers, even if they inform users that they are interacting with chatbots. In Congress, a series of bills have been proposed to ban chatbots from posing as lawyers, doctors, and other licensed professionals. These bills have yet to gain traction.

For now, people will continue to turn to AI to be their lawyer. For many of them, the rewards outweigh the risks. Not long after Judge Braswell asked pro se litigants why they wanted a particular piece of evidence, they would murmur timidly. Now, they confidently answer her questions, having rehearsed with a chatbot.

“It’s a really difficult system to navigate. With AI, however, it becomes a little less complex,” she says.