---
title: The Price of Unverified Caselaw
url: https://justice-denied.org/asiedu-unverified-caselaw-sanctions/
date: 2026-09-09T14:03:59+00:00
modified: 2026-09-09T14:05:21+00:00
lang: en_US
---

# The Price of Unverified Caselaw

A Minnesota appellant used AI-assisted research, failed to verify the cases he presented, and learned that “doing this part time” is not an exception to legal accuracy.

---

Artificial intelligence can help a person find questions, organize an argument, or identify possible starting points for research.

But it cannot make a nonexistent case real.

It cannot turn a real case into support for a proposition the case does not contain. And it cannot take responsibility for a filing after the filing reaches a court.

In *Asiedu v. Eastlund*, the Minnesota Court of Appeals put a dollar figure on that distinction. The court upheld $2,341.25 in costs and attorney fees after a self-represented litigant used artificial intelligence as a research tool and failed to ensure that his caselaw citations were accurate.

The decision is nonprecedential. It does not prohibit AI. It does something more basic: it holds the person who signs and submits a legal document responsible for the authority inside it.

**Case at a glance:** *Asiedu v. Eastlund*, A25-2106 and A26-0532; Minnesota Court of Appeals; filed September 8, 2026. The court affirmed dismissal of the underlying claims on immunity grounds and affirmed a $2,341.25 attorney-fee sanction imposed under the district court’s inherent authority.

![Illustration of a magnifying glass exposing unverified caselaw in a court brief](https://justice-denied.org/wp-content/uploads/2026/09/unverified-caselaw-editorial-1024x576.png)

***Unverified caselaw.** The Asiedu sanctions order underscores that lawyers remain responsible for the authorities they submit, regardless of how the research was produced.*

## A Sanction Built on Failed Verification

Michael Kwabena Asiedu, representing himself, sued Olmsted County and several child-protection employees. His complaint alleged serious misconduct involving the removal of his child from school, the use and dissemination of a safety plan containing false allegations, restrictions on his parental access, and the withholding of records.

The Court of Appeals accepted those allegations as true because it was reviewing a dismissal for failure to state a claim. That procedural assumption is not a factual finding. The court ultimately affirmed dismissal because it held that the individual employees were protected by qualified and official immunity and that the county received vicarious official immunity.

But the part of the opinion that should command wider attention came after the immunity analysis.

The district court found that Asiedu repeatedly misrepresented caselaw in two ways: he cited cases for legal principles they did not support, and he cited cases that did not exist. At the sanctions hearing, he admitted using artificial intelligence to help draft his filings. He also admitted that he had not ensured the citations were accurate because he was “doing this part time.”

The district court found bad faith and ordered him to pay the respondents’ reasonable costs and attorney fees. The total was $2,341.25.

The Court of Appeals upheld that decision.

## Self-Representation Is Not an Accuracy Exemption

Courts sometimes give self-represented parties limited procedural leeway. That does not mean the law becomes optional or that a judge must treat invented authority as a good-faith substitute for research.

The Court of Appeals relied on the familiar rule that self-represented litigants are generally held to the same standards as attorneys when filing documents in court. It also pointed by comparison to [Minnesota Rule of Civil Procedure 11.02](https://www.revisor.mn.gov/court_rules/cp/id/11/). Under that rule, an attorney or self-represented litigant who presents a document certifies, after an inquiry reasonable under the circumstances, that the legal contentions are warranted by existing law or by a nonfrivolous argument for changing the law.

One nuance matters. The district court imposed the sanction under its inherent authority, not through a direct Rule 11 sanctions ruling. The appellate opinion used Rule 11.02 as a comparison for the promise a filer makes when presenting legal contentions to a court.

**The obligation is not to be perfect. The obligation is to conduct a reasonable inquiry and not present confidence as a substitute for authority.**

“I am not a lawyer” can explain why someone needs help understanding procedure. “I am doing this part time” can explain why research takes longer. Neither statement can make a false citation reliable enough for a court to use.

## The More Important Half of the Ruling

Fabricated cases attract attention because they are easy to describe. A citation points to nothing. The failure is obvious once someone checks.

But *Asiedu* addressed a second problem that can be harder to detect: a real case cited for the wrong legal principle.

That form of error looks more legitimate. The case exists. The citation may be correctly formatted. A reader may assume the hard work has already been done. Yet the authority does not support the proposition for which it is offered.

That distinction is central to our earlier article, [“When the Law Hallucinates.”](https://justice-denied.org/when-the-law-hallucinates/) Legal verification is not finished when a database confirms that a case name exists. The person relying on the case must read it closely enough to know what it held, what question it decided, and whether the quoted or summarized language fairly supports the argument being made.

A fake case invents authority. A misused real case borrows authority it does not have. Both ask the court to trust something that has not been verified.

Justice-Denied documented that same problem in [“Distorting the Law – The Common Interest Doctrine (2021)”](https://justice-denied.org/distorting-the-law-the-common-interest-doctrine-2021/), where a proposition attributed to *In re Lawrence* did not appear in the decision and instead came from the defendants’ interrogatory answers.

## From Warning to Monetary Consequence

The public discussion about AI and legal filings often ends with a warning: verify before filing. *Asiedu* shows what can happen when that warning is ignored.

The sanctions were not imposed merely because AI was used. The appellate court focused on repeated inaccurate authority, the failure to check it, the filer’s explanation for that failure, and the district court’s finding of bad faith. The consequence was not abstract. It was an enforceable award of costs and attorney fees.

That matters because the predictable failure mode of generative AI is no mystery. These systems can produce polished language, plausible case names, professional-looking citations, and confident summaries without guaranteeing that any of them are accurate. Anyone who uses the output in court is on notice that fluency is not verification.

## The Rule Should Travel With the Power

The Court of Appeals was right to reject the idea that limited time excuses unverified authority. Courts cannot function if parties are allowed to submit invented cases and leave judges or opponents to discover the fabrication.

But the principle should not stop at the counsel table.

If a self-represented litigant may be ordered to pay thousands of dollars for presenting nonexistent cases or misusing real ones, legal institutions should be equally serious when unsupported authority appears in work carrying greater power. A judicial order can terminate claims, restrict discovery, shape an appellate record, and become the government’s official explanation of why one side lost.

**The duty to verify should rise with the power of the document—not disappear behind it.**

This does not mean every legal mistake is misconduct. Judges, lawyers, and litigants can make honest errors. It means the response should turn on the same core questions: Was the authority actually checked? Does it support the stated proposition? Was contrary language ignored? Was the problem corrected after it was identified? And did the person responsible treat accuracy as an obligation or an inconvenience?

Those questions matter whether the unsupported proposition came from an AI tool, an advocate’s brief, or an official decision.

## The Practical Lesson

The practical rule is simple: use AI as a lead, never as authority.

Open the case. Confirm that it exists. Read the holding. Check the procedural posture. Verify every quotation. Make sure the proposition is supported by the cited page and has not been undermined by later authority. If the source cannot be verified, do not ask a court to rely on it.

That work takes time. Sometimes it takes more time than a self-represented person has. Sometimes it costs money. Sometimes it forces a writer to narrow or abandon an argument.

But accuracy is not the part of legal writing that can be skipped when time runs short—whether by a pro-se litigant, an attorney—or even a judge.

In *Asiedu*, the excuse was that the work was being done part time. The sanction makes the court’s answer unmistakable: the duty to verify applies full time.

## Sources

- [*Asiedu v. Eastlund*, A25-2106 and A26-0532](https://mn.gov/law-library-stat/archive/ctapun/2026/OPa252106-090826.pdf), Minnesota Court of Appeals (filed September 8, 2026).
- [Minnesota Rule of Civil Procedure 11](https://www.revisor.mn.gov/court_rules/cp/id/11/).
- [When the Law Hallucinates](https://justice-denied.org/when-the-law-hallucinates/), Justice-Denied.
- [Distorting the Law – The Common Interest Doctrine (2021)](https://justice-denied.org/distorting-the-law-the-common-interest-doctrine-2021/), Justice-Denied.