The ABA Journal ran a story this week about a man named Daniel Jensen. He is not a lawyer. He is representing himself, and this week he stood up in a Georgia business court and delivered an opening statement to a jury. He got there with the help of artificial intelligence. What happened once he arrived is the most useful thing I have read all year about where this technology actually stands.
I would show you the photograph of him in the courtroom, but it ran with a Law.com story I cannot get to because it is behind a paywall. And I see no one to contact to ask permission to use the photo here. And here is the utter irony…. This is a story about access. It is about people who cannot afford the system trying to find a way in. And the best reporting on it sits behind a wall I tried and failed to get to. I have attempted more than once to buy a Law.com subscription. Their site routes an individual to an email-for-information form. I filled it out and heard nothing back. As far as I can tell, they sell to firms and institutions, not to a single lawyer who wants to read one article. <SHRUG>
So a direct message to ALM, which owns Law.com: your journalism on the access-to-justice gap is good, and I would like to pay you for it. I cannot find a way to do that. There is an irony in reporting on people locked out of the legal system from behind a wall that locks out the individual reader. Tell me how to subscribe and I will. Until then I am linking your reporting below so my readers can at least find it, and citing the underlying study directly so they can read the data even if they, like me, cannot read your article about it.
Start with the accomplishment, because it is real. Jensen filed a breach-of-contract suit, survived to trial in a specialized business court, and stood up to give an opening in front of a judge and jury. Most people never get that far. The large majority of civil claims brought without a lawyer end well before a jury is seated. A self-represented litigant delivering an opening in a specially set jury trial is a real thing, and AI helped put him there.
Now the other half. The same judge presiding over the trial had earlier found thirteen inaccurate or misleading citations in Jensen’s briefs, and ordered both sides to disclose any future use of AI. Jensen’s filings since then disclose that he used Claude and ChatGPT. And during the trial, according to the reporting, he struggled with the things a tool cannot do for him. He asked how to show evidence during witness testimony. He had trouble organizing exhibit copies for opposing counsel and the court. He got himself to the courthouse, and the machine could not walk him through what came next.
I have written before about the line between legal information and legal judgment. Jensen’s trial is that line in physical form. Everything AI did well for him is on the information side. It helped him understand the claim, draft the documents, learn the shape of the process, and prepare an argument. Everything he stumbled on is on the judgment side, which is the actual practice of law. When to object. How to move an exhibit into evidence. How to read a courtroom and adjust in real time. The tool carried him as far as information goes and set him down at the point where judgment begins.
This is bigger than one man. Pro se filings are rising fast, and AI is behind much of the increase. A 2026 working paper from the University of Miami Law and AI Lab found that the federal civil pro se plaintiff rate rose from about 11 percent before generative AI was widely available to about 17 percent after. The people doing this are not chasing a gimmick. They could not afford a lawyer and found a tool that finally let them try. That is the access-to-justice promise showing up in real life, and it should be taken seriously.
There is an irony here worth naming plainly. The promise of these tools is access to justice. The early data shows something more complicated. The same Miami study found that complaints showing signs of AI drafting were dismissed more often than those without, 61.1 percent against 53.6, and it found no improvement in win rates at all.
More people are getting into court. They are not doing better once they arrive, and by some measures they are doing worse. The researchers put their finger on why. These tools are good at producing fluent, citation-dense legal writing. They are less good at building a fact-specific story grounded in the person’s actual circumstances that states a claim the law will recognize. The formatting improves. The case does not.
So the access that AI provides can be its own kind of trap. It gives a person the confidence and the paperwork to walk through the courthouse door, and then leaves them exposed once they are inside, facing a dismissal they might have avoided by never filing that particular claim that particular way. Access to the building is not the same as access to justice. The first is a door. The second is everything that has to happen after you walk through it, and that is the part the technology has not solved.
Both honest arguments belong in the same room. The access argument is true. For someone who cannot afford counsel, a tool that helps is competing with nothing, because nothing is what she had. Telling her to wait for a lawyer she will never be able to hire is not protection. And the caution is also true. A tool that gets a person to a jury trial she is not equipped to run, that fills her briefs with citations that do not exist, and that goes quiet when the practice of law begins, has not delivered access to justice. Both of those are true about the same case.
What Jensen’s trial argues for is not less of this technology, and not more of it uncritically. It argues for a clear understanding of what the technology is for. The tool that helps a person understand her case and prepare is doing real and lawful good. The tool that implies it can substitute for judgment inside a courtroom is setting up the person it is supposed to help to be the one who pays for the gap. Jensen got himself to trial. Whether the technology that took him there can take him the rest of the way is the question the field is going to answer in public now, case by case.
I hope he does well. I also know hoping is not a strategy, and neither is a tool that stops at the courtroom door.
Sources
Amanda Robert, “Pro se litigant starts trial with help of artificial intelligence,” ABA Journal (July 29, 2026). https://www.abajournal.com/news/article/pro-se-litigant-starts-georgia-trial-with-help-of-ai
Original reporting (subscription): Law.com / Daily Report (July 28, 2026), “Pro se clients might use AI to file lawsuits, but can they win them at trial?” https://www.law.com/dailyreportonline/2026/07/28/pro-se-clients-might-use-ai-to-file-lawsuits-but-can-they-win-them-at-trial
Cohen-Sasson, “The New Pro Se: Generative AI and the Surge in Federal Civil Self-Representation,” Miami Law & AI Lab, University of Miami School of Law (working paper, May 28, 2026). https://arxiv.org/abs/2605.29493 (dismissal rates of 61.1% vs. 53.6%; no improved win rate)
“Big Law Grapples With AI-Fueled Pro Se Surge, Rising Legal Costs,” Bloomberg Law (Mar. 12, 2026). https://news.bloomberglaw.com/business-and-practice/big-law-grapples-with-ai-fueled-pro-se-surge-rising-legal-costs
“Pro Se Workers’ AI Assists Lead Courts to Warnings, Sanctions,” Bloomberg Law (Apr. 10, 2026). https://news.bloomberglaw.com/daily-labor-report/pro-se-workers-ai-assists-lead-courts-to-warnings-sanctions