Harvey v. Torrent Leasing & U.S. Bank
U.S. District Court, District of Nevada · D. Nev. · Nevada bar guidance
Conduct
Pro se Section 1983 plaintiff cited a fake quote from an overruled Seventh Circuit case and miscited a Nevada case to a different opinion.
Consequence
No AI-citation sanction imposed; case dismissed on the merits (federal claim with prejudice, state claims without).
Lesson
Even when AI defects do not change the outcome, Dorsey will name ChatGPT specifically and document the verification standard.
Verified May 7, 2026
- Citation
- Harvey v. Torrent Leasing & U.S. Bank, No. 2:25-cv-00824-JAD (D. Nev. Dec. 15, 2025) (Dorsey, J.)
- Decided
- December 15, 2025
Summary
Pro se plaintiff Harvey sued Torrent Leasing and U.S. Bank in the District of Nevada (No. 2:25-cv-00824-JAD) on a Section 1983 claim with pendent state-law theories. In a December 15, 2025 dismissal order, U.S. District Judge Jennifer A. Dorsey identified two citation defects: a fake quote attributed to the (overruled) Seventh Circuit case Nesses v. Shepard, and a miscited "Flangas v. Perfekt Marketing" whose citation actually corresponded to Lawrence v. Clark County. Dorsey expressly named ChatGPT in cautioning that "generative AI software like ChatGPT... may hallucinate (or make up fake) legal authority" and warned that "using generative AI tools like ChatGPT may result in sanctions" without independent verification.
- AI tool:
- ChatGPT (court named ChatGPT specifically as exemplar of generative AI that "may hallucinate")
What sanction did the court impose?
No monetary sanction imposed for the AI-citation defects. On the merits, U.S. Bank's motion to dismiss was granted: the Section 1983 claim was dismissed with prejudice; the state-law claims were dismissed without prejudice under 28 U.S.C. § 1367; the default entry against Torrent Leasing was vacated; the case was closed in defendants' favor.
Why does Harvey v. Torrent Leasing & U.S. Bank matter for law firms using AI?
Harvey is the December 15, 2025 entry in Judge Jennifer A. Dorsey’s two-order AI-citation pattern in D. Nev. (this case plus the August 11, 2025 Chavez-DeRemer v. NAB order). Both orders use the same “hallucinate (or make up fake) legal authority” framing and both explicitly name ChatGPT. Dorsey’s chambers are now among the most named-AI-tool-explicit in the federal AI-sanctions landscape.
The fabrication pattern here is the harder one to detect. Harvey’s brief attributed a fabricated quotation to Nesses v. Shepard, an overruled Seventh Circuit case. The citation itself resolves to a real opinion. Only reading the cited text reveals the defect. AI-verification policies that stop at confirming the cite shepardizes will miss this pattern entirely; the verifying associate has to pull and read the cited authority.
Dismissal of the case on the merits was independent of the AI-citation issue. Dorsey nonetheless documented the AI defects in the order. That practice builds a chambers-level record for any future Rule 11 escalation.
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.
- Note Judge Dorsey's chambers explicitly names ChatGPT as the prototypical AI tool and uses the 'hallucinate' framing; this is a chambers-level signal that AI-citation issues are being monitored on every filing.
- When defending a Section 1983 case in D. Nev. before Dorsey, document AI-citation defects in the dismissal motion itself even if they do not affect the merits outcome; Dorsey's chambers will incorporate the documentation into the order.
- The 'fake quote from a real (overruled) case' pattern is harder to spot than the fully fabricated case; require the verifying associate to read the cited opinion in full, not just confirm the citation resolves to a valid case.