AI-Drafted Court Filings

When Clients Turn to AI and Their Cases Go Sideways

These days, I often meet clients who tried handling their own lawsuits without a lawyer, ran into trouble, and only then found my blog and reached out. Against that backdrop, a recent article in the Chosun Ilbo's “Seocho-dong 25 Si” column caught my attention, because it captured exactly what I've been seeing firsthand. Here's a summary of what the article reported.

According to the report, a district court judge recently found the phrase “according to Gemini” left in a loan-repayment complaint — a stray trace of the AI tool used to draft it. Judges say this is becoming routine, and one presiding judge described the experience as feeling like a case between plaintiff “ChatGPT” and defendant “Gemini.” Untangling what a party actually meant, then rewriting the record, can eat up days of a court's time.

— Yang Eun-kyung, “Plaintiff ChatGPT, Defendant Gemini? Judges Struggle With AI-Written Filings,” Chosun Ilbo, Seocho-dong 25 Si, August 14, 2026.

The Moment AI Authorship Shows, Credibility Collapses

Once it becomes clear that a filing was written by generative AI, the court's trust in that filing tends to drop fast. And the problem doesn't stop there. To save on legal fees, more and more parties will likely be tempted to use AI to produce sprawling filings that run dozens of pages. But Korean courts don't respond well to filings that are overly long or that simply repeat the same argument in different words — if anything, that kind of filing just wears the court down. Experienced lawyers know this, which is why they pick out only the points that matter and argue them concisely.

Self-represented parties, in particular, are turning to AI to produce filings that run dozens of pages — partly to save on legal fees, and partly because it's tempting to keep repeating the argument they want to make in different words. Even when the court specifically asks them to clarify a point, many respond by restating their own position rather than actually answering the question. That approach can end up hurting their case rather than helping it.


Witness Examination Is About to Matter More Than Ever

I expect that in higher-stakes cases, witness examination will increasingly become a required step. As AI makes it easier to fabricate not just written arguments but evidence itself in convincing ways, courts will find it harder to rely on filings and documentary evidence alone to establish the facts.

But catching a key statement in real time during witness examination, spotting a contradiction in the other side's answer, and following up on the spot — that's a skill built from years of courtroom experience. It's not something an ordinary person can pick up quickly, and you obviously can't pause mid-examination to ask an AI what to say next. So the more a case turns on disputed facts, the more real value a lawyer brings to it.

Fines for AI Misuse Should Apply to Parties, Not Just Lawyers

At this point, I think we genuinely need some way to flag whether a filing was drafted by ChatGPT or another generative AI tool. Courts need at least a basic mechanism for judging which filings can be trusted — otherwise the burden on judges keeps growing, and the reliability of the whole process suffers.

The plan recently announced by the National Court Administration would let courts fine parties who cite fabricated case law, and refer lawyers who do the same to the Korean Bar Association for discipline. But as a practical matter, that kind of sanction is hard to enforce against self-represented litigants who never hired a lawyer in the first place. For the system to actually be balanced, ordinary parties who flood the court with AI-generated filings need to face the same kind of fines. I'll be honest — I find it hard to understand why courts would fine only the lawyers, who tend to behave professionally, while letting the parties themselves, who can sometimes become genuinely difficult litigants, off the hook. As it stands, it feels like the system in Korea penalizes the very people who follow courtroom norms, which is a strange, self-defeating way to run a court. If parties who overuse AI while representing themselves actually faced fines, they'd start taking the decision to hire a lawyer more seriously — and that, in turn, would naturally help restore order to the litigation process.

A Culture of Hiring Lawyers Is What Keeps Courts Orderly

Whenever I make this argument, I get pushback along the lines of, “Isn't that just good for business if you're a lawyer?” But I genuinely believe that a culture where people hire lawyers matters for keeping the courts running smoothly. Under the current setup, where self-represented parties can freely submit sprawling, unorganized filings, judges end up with a much heavier workload, and legally unrefined arguments pile up in a way that can throw off the order of the whole proceeding. And the time that gets wasted doesn't just affect that one case — it ripples out to every other litigant waiting for their turn before the same court. Building a stronger culture of hiring lawyers isn't about protecting any one profession's interests. It's about using judicial resources efficiently and protecting the integrity of the litigation process as a whole.

Closing Thoughts

Generative AI is now an irreversible part of how litigation works in practice. My hope is simply that this trend doesn't end up eroding trust in the judicial process — which means we need both institutional safeguards and a clearer understanding of what lawyers actually bring to the table.


Post a Comment

Previous Post Next Post