Lepp v. Mallett
U.S. District Court, Eastern District of Michigan · E.D. Mich. · Michigan bar guidance
Conduct
Pro se Plaintiff attached an AI Legal Analysis Report from ChatGPT-4o that misrepresented Lukumi as a standing case.
Consequence
Rule 60(b) motion denied; AI exhibit declined for reliability concerns; no sanctions but explicit Rule 11 warning.
Lesson
Levy treats AI-generated 'legal analysis reports' as unreliable exhibits; accuracy review survives even candid AI disclosure.
Verified May 7, 2026
- Citation
- Lepp v. Mallett, No. 5:25-cv-10214 (E.D. Mich. Jan. 16, 2026) (Levy, J.)
- Decided
- January 16, 2026
Summary
Pro se Plaintiff Heidi Grossman Lepp filed a motion for relief from judgment under Fed. R. Civ. P. 60(b) seeking to vacate dismissal of her First Amendment church-discrimination case. Attached to her proposed objections was an "AI Legal Analysis Report" that Plaintiff disclosed was generated by "OpenAI's ChatGPT-4o model" to perform what she described as an "objective and dispassionate review" of the magistrate judge's report and recommendation. The report characterized Church of the Lukumi Babalu Aye v. City of Hialeah, 508 U.S. 520 (1993), as supporting Plaintiff's standing to assert rights of her congregation, when in fact Lukumi does not address standing.
- AI tool:
- OpenAI ChatGPT-4o (Plaintiff disclosed using ChatGPT-4o to produce an 'AI Legal Analysis Report' attached as an exhibit)
What sanction did the court impose?
Motion for relief from judgment denied. District Judge Judith E. Levy declined to consider the AI Legal Analysis Report exhibit on reliability grounds, finding the document "inappropriately presented and unhelpful." Court declined to impose sanctions but catalogued the "Report's" inaccuracies and emphasized that pro se litigants who rely on generative AI without reviewing accuracy of citations risk Rule 11 sanctions. Judgment stands.
Why does Lepp v. Mallett matter for law firms using AI?
Lepp v. Mallett is a useful counterpoint to the Patti chambers’ Ali-framework cases. Where Patti tends to issue direct warnings tied to the $200-per-citation framework, Judge Levy’s approach in Lepp is to treat the AI exhibit as a reliability problem and decline to consider it on the merits, rather than impose sanctions. The doctrinal grounding is the same Hunt v. Morissette / Sanders v. United States / Evans v. Robertson stack that Patti uses, but Levy’s procedural posture is different: she does not order the AI material stricken, she just refuses to consider it, while leaving open the possibility of sanctions in a future filing.
The case is also instructive on the limits of AI disclosure as a defense. Plaintiff openly disclosed that the “AI Legal Analysis Report” was produced by ChatGPT-4o and described it as offering “an objective and dispassionate review free from advocacy, emotion, or bias.” Levy was not persuaded: the order walks through specific inaccuracies in the Report, including the misrepresentation of Church of the Lukumi Babalu Aye as a standing case (it is a free-exercise case that does not address standing). The implication for litigants and counsel is that voluntary disclosure of AI use does not preempt accuracy review; the court will read the AI output and decline to consider material that contains errors regardless of how candid the disclosure was.
For E.D. Michigan firms representing First Amendment plaintiffs or any pro se litigant inclined to attach AI-generated “legal analysis,” the practical takeaway is that the practice creates a documentary record of unreliable assertions that the court can use to deny relief on the merits. Counsel taking over a pro se case mid-stream should remove or refuse to incorporate any AI-generated analytical exhibits before filing further motions.
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.
- Disclosure that an exhibit is AI-generated does not insulate the filing from substantive scrutiny; courts will read the AI output for accuracy and decline to consider it if the analysis is wrong.
- Religious-liberty plaintiffs and their counsel should not present AI-generated case analyses as supplemental persuasive authority; the practice triggers reliability scrutiny that ordinary briefing does not.
- Levy's order cites Hunt v. Morissette, Sanders v. United States, and Evans v. Robertson as the doctrinal stack; firms responding to AI-generated filings can cite this triad as established E.D. Mich. authority.