Skip to content
Regulayer™Human Control for AI
Book a live demo

The sanctions record · United States District Court, Eastern District of Michigan, 2026

Chirece C. v. Commissioner of Social Security

Chirece C. v. Commissioner of Social Security, Case No. 25-11554 · Order Fining Plaintiff’s Counsel for Misuse of Artificial Intelligence, Doc. 20 · Magistrate Judge Elizabeth A. Stafford · dated 24 September 2026

Order read in full, seven pages; every quotation is from it. Source: the order as filed, obtained as a PDF from the AI Hallucination Cases document store.

A fine with a deadline, and a warning about every brief still uncorrected. In a Social Security disability appeal, the court found two citations in the plaintiff’s summary judgment brief that suggested AI use. Counsel admitted using AI and filing without verifying them. She had already been sanctioned two months earlier in another case in the same court. The court fined her $500, payable to the Clerk by 16 October 2026, and warned her and her firm that “more significant sanctions will be imposed” if briefs with fake citations remain uncorrected after that date. A footnote adds a second concern: her firm’s use of AI to summarize medical records. “This is concerning.”

The record

  • CourtUnited States District Court, Eastern District of Michigan, Southern Division
  • JudgeElizabeth A. Stafford, United States Magistrate Judge, conducting all proceedings with the parties’ consent
  • CaseNo. 25-11554, an appeal from a partially favorable decision on disability insurance benefits and supplemental security income
  • Dated24 September 2026
  • SanctionedPlaintiff’s counsel, Erin Rich. She and her firm, Disability Law Group, were warned
  • The errorTwo citations in the summary judgment brief, identified by the court itself. The court discovered them on 3 September 2026
  • ToolAI, not named. Use admitted in the response to the order to show cause, with the admission that the citations were not verified
  • AmountA fine of $500.00, payable to the Clerk of the Court by 16 October 2026
  • WarningMore significant sanctions if briefs with fake citations, factual misrepresentations or misleading holdings remain uncorrected after 16 October 2026
  • EarlierFarley v. Comm’r of Soc. Sec., No. 26-10137 (E.D. Mich.), 13 July 2026: a Rule 11 sanction against the same counsel and firm, four continuing legal education courses and a review of pending motions, no monetary sanction

What the court found

The court set the case against what it called the exploding use of generative AI in litigation, citing the Sixth Circuit’s holding in Whiting v. City of Athens that even a single fake case can be sanctionable:

“The explosive misuse of AI in court filings is serious matter that must be met with weighty consequences.”

Counsel argued that she should not be sanctioned again so soon after Judge Patricia T. Morris’s order in Farley. The court did not accept that. The citations in this brief “remained uncorrected here for almost two more months” after the Farley order, until the court found them. “And the number of Rich’s filings that continue to include fake AI generated information remains unknown to this Court.” In another of her cases before the same judge, Host v. Comm’r of Soc. Sec., No. 25-13451, a cited “Williams” decision does not exist, and two real decisions were cited for rules they do not state.

Her firm had put new internal policies in place. “But that new policy does not obviate her responsibility to swiftly correct existing filings that include AI generated false information.”

Footnote 2 turns from citations to client files. Counsel had told Judge Morris that her firm “occasionally uses AI to summarize voluminous medical records.” The court:

“This is concerning. See ABA Model Rules of Pro. Conduct Rule 1.6 (confidentiality of client information); Mich. Rules of Pro. Conduct Rule 1.6 (same)”

It also quoted a Social Security practice handbook: “Uploading medical records for review and summary, problematic. To do so, requires the provision of unredacted client records containing their most sensitive health information to a service that openly concedes it may use that data for training.”

Why it matters

Three features travel. First, the deadline. The fine carries a date, and so does the warning, which speaks to any of her briefs that still hold fake citations after 16 October 2026, not only the one before the court.

Second, the repeat. A sanction in July in one case did not end it; the uncorrected brief in this case drew a fine in September. A firm policy adopted after the first sanction did not answer for filings made before it.

Third, the footnote. A federal court has flagged, citing the confidentiality rule, the practice of putting clients’ medical records into an AI service to summarize them. That is a question about what goes into the tool, not only about what comes out of it.

The lesson for the next filing

The court could not say how many other filings still carried fake citations, and neither, on this record, could counsel. A record made at the time of each filing, of what was checked and by whom, is the only way to answer that question brief by brief.

Regulayer™ creates checkable evidence of who approved exact words and when: which AI was used, what a named person reviewed, and the document it was sealed to, made at the time and verifiable free by the court without trusting the person who made it. It does not do the checking. It shows that the checking happened, and who stood behind it.

Verify a sealed record free →

Sources

Related

Checked against the order, read in full. Information, not legal advice.

Information, not legal advice. Every entry is verified against the court’s own document or contemporaneous reporting; where a source is reporting rather than the document, we say so.