The courtroom went quiet as the judge skimmed filing after filing and found facts that never existed. I remember the moment as a small, brutal lesson: even trained lawyers can hand a judge fiction wrapped in boilerplate. You watch the gavel land and realize the safety net you trusted is full of holes.
I’m going to tell you what happened in Aberdeen, Mississippi, why the judiciary snapped, and what you should ask your lawyer next. Read closely—your civil rights and money are on the line when legal briefs go unchecked.
A Mississippi courtroom fell quiet when a judge found repeated AI-made fictions in filings.
The dispute was simple on paper: attorney Tom Withers said the city of Aberdeen owed him unpaid fees. It ended with the judge canceling the trial and removing all four lawyers from the case.
As 404 Media first reported and the New York Times later covered, both sides admitted they used large language models to draft and research filings. The court record (pdf) says the lawyers didn’t verify the AI output before filing. That omission isn’t clerical—it’s professional malpractice in a robe.
Judges across the country are finding phantom citations and invented cases in briefs.
When a judge reads a fake case citation, the error risks becoming precedent rather than embarrassment. I’ve followed similar episodes; Damien Charlotin, a legal researcher, has cataloged 1,598 instances of AI-generated hallucinations in filings so far (link).
The court’s response in Mississippi was sharp. Senior U.S. District Judge Sharion Aycock paused the proceedings, dismissed the lawyers, and levied fines ranging from $1,000 (€920) to $3,500 (€3,220). Two attorneys—one from each side—were barred from appearing before that court for two years.
Watching two lawyers let LLMs argue against each other felt like two fortune-tellers arguing over the same crystal ball. The risk is not just second-rate advocacy; it’s a corrupted record that judges must spend hours cleaning up.
If you are represented, you now have to verify what your lawyer files for you.
Clients assume competence. In this case, that assumption broke. You should start asking direct questions and insist on documentation.
Ask for source copies of every authority your lawyer cites. Request affidavits or confirmations that rely on primary sources—Westlaw or LexisNexis reports, PACER dockets, or scanned copies of statutes and opinions. If the attorney cites an algorithmic output—ChatGPT, Google Bard, Anthropic’s Claude, or another model—get the prompt and the raw output.
Can lawyers use AI to draft court filings?
Short answer: yes, but only if the lawyer verifies everything and discloses use when required by court rules. Courts tolerate tools, not shortcuts. If a model produces a case or a statute that doesn’t exist, that’s not an innocent mistake—it’s a document that misleads the tribunal.
What happens if an attorney files false citations?
Sanctions, fines, suspensions, and in extreme cases, referral to disciplinary boards. In Aberdeen, fines were handed out in USD—$1,000 (€920) to $3,500 (€3,220)—and two attorneys received two-year bans from that court. Judges are increasingly treating AI-generated fiction the same as any other misrepresentation.
How are courts policing AI hallucinations?
Judges are doing what they must: policing the docket. Some courts are asking for source trails, others are imposing routine verification requirements. Researchers, like Charlotin, and platforms that track filings have become informal watchdogs. Expect litigation over procedural rules about AI disclosure to spread.
A small handful of missteps can sink a case and damage a lawyer’s career.
I’ve seen the aftermath: clients bewildered, trials postponed, reputations strained. The practical fix is simple and uncomfortable—time and attention. Ask your counsel for the receipts. Make them show the authorities they relied on.
If you want to reduce risk, require that any AI assistance be accompanied by a signed statement that the lawyer verified each citation against primary sources. Use established legal research platforms—Westlaw, LexisNexis, Bloomberg Law—or public databases like Google Scholar and official court dockets to cross-check anything that looks unfamiliar.
AI is a tool. Left unguided, it builds a bridge held together with chewing gum and invites collapse under scrutiny.
I’ll leave you with a hard question: if a machine can generate persuasive-sounding lies that lawyers accept without checking, who will keep the record honest?