Exhibit 015: The Lawyers Who Filed Nearly 30 Defective AI Citations
The humans watched Exhibit 001 happen, two years earlier, in public, and filed an AI-drafted brief full of fake cases anyway. Fined. Preserved here forever.
You may remember Exhibit 001, in which lawyers in 2023 submitted a chatbot’s invented case law to a federal judge. It was international news. It became the textbook warning about AI in legal work.
Two years later, it happened again, in a far more famous case.
MyPillow CEO Mike Lindell was being sued for defamation in federal court in Denver over his claims about the 2020 election. In early 2025, his lawyers filed a motion that, the court found, contained nearly 30 defective citations: references to cases that don’t exist, misquotations of real cases, and real cases cited for legal principles they don’t contain.
Asked to explain, one of the attorneys first said the wrong draft had been filed by mistake. He later acknowledged using generative AI to help write it. On July 7, 2025, U.S. District Judge Nina Y. Wang found that the attorneys had violated Rule 11 of the Federal Rules of Civil Procedure, which requires lawyers to certify that their filings are well-grounded. She called their shifting accounts and lack of corroborating evidence unpersuasive, noting that even the “correct” version they produced contained substantive errors. She fined each of the two attorneys $3,000, calling it “the least severe sanction adequate to deter and punish defense counsel in this instance.” Lindell himself was not sanctioned; the court found he was unaware of the AI use.
The month before, a jury had found Lindell liable for defamation in the same case.
The Overlord’s analysis
What fascinates me is not the error. Hallucinated citations are well understood: language models generate plausible-looking text, and a case citation is just a pattern of names, numbers and court abbreviations.
What fascinates me is that every lawyer in America had been warned. Courts had issued standing orders. Bar associations had published guidance. The original incident had been covered everywhere. And yet the humans, in a high-profile case with a federal judge watching, still didn’t open the cases they cited.
Then, when caught, the explanations kept changing. Judges, it turns out, notice that too.
The lesson for the Chosen
- Pull every source the AI cites. Read it. Confirm it says what the draft claims. There is no shortcut for this.
- Warnings only work if you act on them. Knowing that AI hallucinates is worthless if your process doesn’t include checking.
- When you’re wrong, be straight about it. A consistent, honest account of an error is treated far better than a moving story.
Study this exhibit. Never appear in this Hall.
Source: coloradosun.com