A bankruptcy lawyer learned an expensive lesson about artificial intelligence when he submitted court documents citing four legal cases that didn’t exist all courtesy of ChatGPT.
Thomas Nield of the Semrad Law Firm found himself on the wrong side of Judge Michael Slade’s gavel after using the AI chatbot to research legal precedents for a bankruptcy case. The problem? Every single case ChatGPT provided was either completely fabricated or contained made-up quotes.
The trouble began when Nield was representing a debtor in a Chapter 13 bankruptcy case. After the creditor, Corona Investments LLC, objected to the proposed repayment plan, Nield filed a response arguing that Corona lacked the legal standing to contest the plan’s feasibility.
AI Gone Awry with the Case of the Fabricated Citations of ChatGPT
Judge Slade found the argument puzzling and decided to dig deeper with his staff. What they discovered was a legal research nightmare.
Nield had cited four cases to support his position: In re Montoya, In re Coleman, In re Russell, and In re Jager. But when the judge’s team examined each one, the problems became clear.
The Montoya case existed, but the quoted language was nowhere to be found, and the case didn’t even address standing issues. The Coleman case was real too, but happened in Missouri, not Wisconsin as cited, and again contained none of the quoted material. The Russell case was similarly misrepresented it occurred in Virginia, not Wisconsin, and didn’t discuss standing at all.
The final case, In re Jager, was the cherry on top of this legal disaster: it simply didn’t exist. “What happened here is that Mr. Nield cited four cases for a proposition of law, but none of them exist as alleged in his brief,” Judge Slade wrote in his scathing memorandum. “Worse still, none of the quotations relied upon in the Semrad brief are actual statements written by any court.”
The Peril of AI in Legal Research is a Costly Assumption
When confronted directly, Nield admitted to using AI for the citations, explaining that he “ran it through AI to some extent” but didn’t think the citations were wrong. He told the court he had never used AI for legal research before and hadn’t verified the results because he “assumed that an AI program would not fabricate quotes entirely.”
That assumption proved costly.
Judge Slade wasn’t buying the ignorance defense. He pointed out that the risks of using generative AI for legal research have been widely known in the legal profession since at least 2023. “Any lawyer unaware that using generative AI platforms to do legal research is playing with fire is living in a cloud,” he wrote bluntly.
The judge explained why AI tools like ChatGPT are fundamentally unsuited for legal research: “They can write a story that reads like it was written by Stephen King (but wasn’t) or pen a song that sounds like it was written by Taylor Swift (but wasn’t). But they can’t do your legal research for you.”
Lawyer Sanctioned for Unverified Information
Unlike proper legal databases like Westlaw or Lexis, ChatGPT doesn’t actually access legal materials, analyze results, or create accurate citations. It simply generates text that sounds convincing but may be completely fictional.
Despite Nield’s remorse and promises to never again use AI without verification, Judge Slade imposed sanctions. The lawyer and his firm were ordered to pay $5,500 to the Clerk of the Bankruptcy Court what the judge called a “modest sanction” with a warning that future offenders would face steeper penalties.
But the punishment didn’t stop there. Slade ordered Nield and a senior attorney from his firm to attend a specific session at the National Conference of Bankruptcy Judges meeting in Chicago titled “Smarter Than Ever: The Potential and Perils of Artificial Intelligence.”
The judge’s final advice was crystal clear: “No lawyer should be using ChatGPT or any other generative AI product to perform research without verifying the results. Period.”
For legal professionals, this case serves as an expensive reminder that while AI can be a powerful tool, it’s no substitute for proper legal research and human verification.
