Jarrus v. Governor of Michigan
U.S. District Court, Eastern District of Michigan · E.D. Mich. · Michigan bar guidance
Conduct
Pro se Plaintiffs filed AI-generated objections containing three misrepresented citations to Krupski, Heller, and Oppenheimer.
Consequence
$600 total ($300 per Plaintiff) under Rule 11; reconsideration denied; further AI misuse risks dismissal or IFP revocation.
Lesson
Behm scaled Patti's $200/citation to $300 after Plaintiff's ChatGPT Plus subscription defeated the IFP affordability defense.
Verified May 7, 2026
- Citation
- Jarrus v. Governor of Michigan, No. 4:25-cv-11168 (E.D. Mich. Dec. 2, 2025) (Behm, J.)
- Decided
- December 2, 2025
Summary
Pro se Plaintiffs Michael D. Jarrus and Linda Jarrous filed objections to a magistrate-judge order that contained AI-generated artifact language ("Here's the revised Paragraph 2, ...") and three misrepresented citations to real cases: Krupski v. Costa Crociere S. p. A., 560 U.S. 538 (2010) (misrepresented as making relation-back amendment mandatory); District of Columbia v. Heller, 554 U.S. 570 (2008) (misrepresented as applying Elrod-style irreparable-harm presumption to Second Amendment); and Oppenheimer Fund, Inc. v. Sanders, 437 U.S. 340 (1978) (cited for Rule 26 discovery scope despite being superseded by statute). Plaintiffs filed the AI-assisted objections in the same document in which they objected to Magistrate Judge Anthony P. Patti's prior warning about generative AI use.
- AI tool:
- ChatGPT (Plaintiff Michael Jarrus admitted using ChatGPT Plus, $20/month subscription)
- Sanction amount:
- $600 ($300 per Plaintiff)
What sanction did the court impose?
Three orders document the case. (1) On 2025-11-07, Magistrate Judge Patti issued an omnibus order (ECF No. 168) directing Plaintiffs to read Fed. R. Civ. P. 11 and warning that any AI-generated phantom citation would be sanctionable, citing the $200-per-violation framework from Ali v. IT People. (2) On 2025-12-02, District Judge F. Kay Behm issued an Opinion and Order (ECF No. 176) imposing $300 per Plaintiff ($600 total, due February 2, 2026) for three misrepresented AI-generated citations. The court adopted Patti's $200-per-citation framework and scaled it to $300 after Michael Jarrus disclosed a $20/month ChatGPT Plus subscription that established ability to pay despite his in forma pauperis status. (3) On 2026-01-05, Judge Behm denied Plaintiffs' motion for reconsideration (ECF No. 186), rejecting Michael Jarrus's proposal to reduce his sanction to $60 or set up a $20/month payment plan.
Why does Jarrus v. Governor of Michigan matter for law firms using AI?
Jarrus v. Governor of Michigan is the most thoroughly documented AI-sanctions case from the Behm/Patti pairing on the E.D. Michigan bench, and it is the case where the Ali $200-per-citation framework was scaled and re-applied to a different defendant. The fact pattern is unusual: Plaintiffs filed AI-assisted objections to the very order that warned them about generative AI, with the AI’s chat-interface artifact language (“Here’s the revised Paragraph 2, …”) visible in the filing. Patti’s omnibus order on 2025-11-07 was the warning shot; Behm’s December 2 order quantified the harm at $300 per Plaintiff after Plaintiff Michael Jarrus admitted to a $20/month ChatGPT Plus subscription, which the court treated as evidence of ability to pay despite IFP status.
The reconsideration order on 2026-01-05 is the doctrinally interesting piece. Judge Behm rejected Michael Jarrus’s argument that his sanction should be reduced to a nominal $60 or paid out in $20/month installments, on the rationale that Jarrus had been specifically warned by Patti that financial sanctions would be incurred at $200 per AI-generated phantom citation. The reconsideration order also forecloses the “AI as assistive technology” framing: the court did not bar Jarrus from using ChatGPT, but ruled that a downstream effect on his subscription budget did not justify modifying the sanction. For E.D. Mich. practitioners, the practical takeaway is that the Ali/Jarrus pairing now establishes a per-citation pricing regime for AI hallucinations, with the court willing to scale the figure based on disclosed AI-tool subscription expenditures.
The Linda Jarrous strand of the order is also worth noting. Linda Jarrous submitted a declaration stating that she relied on her son Michael to draft the documents she signed, reviewing them “only at a very general level.” Judge Behm held that Rule 11 imposes obligations on each signatory independently, and that Linda Jarrous would be held equally responsible for the misrepresented citations going forward. A second instance of her failure to verify could result in dismissal of her claims in their entirety. Firms drafting joint filings for related parties should treat each signatory’s verification obligation as independent and document each party’s review separately.
Implications for your firm
Operational steps a firm reading this case may wish to consider documenting. Strategic and rule-application calls belong to your firm's attorneys.
- Treat misrepresentations of real cases (extending a holding beyond what the case actually says) as Rule 11 violations on objective-reasonableness grounds, regardless of subjective good faith.
- When advising pro se litigants who claim inability to pay sanctions, document any AI-tool subscription expenditures; courts now scrutinize them as evidence of ability to pay.
- A signing party who reviews briefs 'only at a very general level' is independently liable under Rule 11; co-signers cannot delegate verification.