A Minnesota Attorney Filed AI-Hallucinated Cases and Paid with 30 Days of His Law License
What happened
In 2025, Minnesota attorney Faisal S. Ahmed filed a legal memorandum in a Hennepin County civil case that cited cases that did not exist. The authorities were AI-generated, and apparently neither Ahmed nor anyone else checked them against an actual legal database before the document reached the court. One citation quoted a real decision, but inaccurately. The rest were invented outright.
The errors did not go unnoticed. Opposing counsel had to track down the phantom citations to confirm they were not real, and the Hennepin County District Court had to spend time establishing the same thing. Courts operate on the foundational assumption that an attorney signing a brief has verified the sources in it. When fabricated references appear in a filed document, the burden falls on everyone else in the room to prove a negative.
Ahmed was fined $5,000 by the court. He subsequently agreed to a 30-day suspension of his law license and an additional $900 in costs. Together, the sanctions reflect a determination that submitting unverified AI output as legal authority is not a clerical lapse but a failure serious enough to interrupt a lawyer's ability to practice.
The incident fits a pattern that has now repeated itself in courtrooms across the country. The underlying mechanics are consistent each time: an attorney relies on a tool that produces fluent, citation-shaped text without any mechanism for distinguishing real case law from plausible-sounding invention. The output looks like research. Nothing in the interface distinguishes "resembles a real case" from "is a real case," and nothing flags when the model fills a gap with a fabricated authority rather than leaving it empty.
What remains absent in each of these cases, and what every court involved has had to reconstruct from the damage trail, is any record of what the tool actually produced, when it was used, or whether a verification step took place before filing. A provable record of what a system did, and whether any check ran between generation and submission, would shift that accountability upstream: surfacing the fabrication before it reaches a judge or opposing counsel, rather than requiring someone else to do that work after the fact.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.