In the Matter of Brian E. Mitchell
U.S. Patent and Trademark Office (Director; Office of Enrollment and Discipline) · USPTO
Conduct
Here a patent litigator used two AI tools on a claim construction chart whose citations to the patent's own record were wrong or nonexistent.
Consequence
Here the USPTO publicly reprimanded him by settlement for competence, diligence, and misrepresentation violations, though the court had not sanctioned him.
Lesson
Here the USPTO warned that AI citation errors are not limited to case law and may reach a patent's specification, figures, and file history.
Verified September 28, 2026
- Citation
- In re Mitchell, Proceeding No. D2026-16, Final Order (USPTO July 27, 2026)
- Decided
- July 27, 2026
Summary
Brian E. Mitchell, a San Francisco patent attorney registered with the USPTO since 2002 who had never prosecuted a patent or trademark application, represented the patent owner in Magpul Industries Corp. v. Mission First Tactical Group, Inc., No. 2:24-cv-5551 (E.D. Pa.). Under the stipulated facts in the USPTO's final order, when preparing the plaintiff's portion of a Joint Claim Construction Chart he used one generative AI tool to help draft proposed claim constructions and a second AI tool to review the result. The AI-generated sections cited the patent's intrinsic evidence, including the specification, figures, and prosecution history, and "[a]t least some of the AI-generated citations were erroneous." The parties to the litigation then determined that there were substantial errors in the plaintiff's citations to the intrinsic evidence: "many of the plaintiff's citations, quotations, and parentheticals referred to portions of the record that did not exist or were inaccurately attributed." When the errors were identified, he reviewed the chart and acknowledged additional erroneous citations, and a revised claim chart was circulated the following day with accurate citations. He acknowledged the citations were not sufficiently checked against the source documents. The district court did not sanction him or his client, and, he represents, the case later settled and was dismissed with prejudice.
- AI tool:
- Two generative AI tools, not named (per the stipulated facts, one to draft proposed claim constructions and a second to review the plaintiff's portion of the chart)
- Sanction amount:
- Public reprimand by settlement; publication in the OED FOIA Reading Room and a notice in the Official Gazette; each side bears its own costs
What sanction did the court impose?
The USPTO Director, acting through Associate General Counsel Tricia Choe on delegated authority, approved a settlement agreement between Mitchell and the OED Director and entered a final order publicly reprimanding him. Mitchell acknowledged violating 37 C.F.R. 11.101 (competence), 11.103 (diligence), 11.804(c) (conduct involving misrepresentation), and 11.804(d) (conduct prejudicial to the administration of justice). The stipulated conclusions include not fully understanding the inherent risks of AI research and drafting tools, including hallucinated citations or quotations; not adequately reviewing citations to the intrinsic record; and presenting inaccurate support, including false citations and quotations, to the district court. As mitigation the order lists prompt acknowledgment and correction, acceptance of full responsibility, no prior discipline, full cooperation with OED, and genuine contrition. The order directs publication in the OED Reading Room and a notice in the Official Gazette, which cautions practitioners that AI citation errors may also extend to the intrinsic evidence of patent and trademark applications. Mitchell waived reconsideration, review, and appeal.
Why does In the Matter of Brian E. Mitchell matter for law firms using AI?
This order is a regulator’s response to AI errors that drew no court sanction. In the Eastern District of Pennsylvania patent case, the errors appeared in citations to the patent’s own intrinsic record rather than to case law, were corrected the next day, and drew no sanction. The USPTO’s Office of Enrollment and Discipline nonetheless pursued the matter, and Mitchell agreed to a public reprimand and acknowledged four violations of the USPTO Rules of Professional Conduct, including 37 C.F.R. 11.804(c) on misrepresentation.
The stipulated conclusions frame the competence and diligence duties in AI-specific terms: not fully understanding “the actual or potential inherent risks (e.g., mistakes, hallucinated citations or quotations, etc.) of AI-based legal research and drafting tools,” and not “making reasonable use” of those tools. The published notice goes further, reminding practitioners of their duties under 37 C.F.R. 11.18, which “in relevant part, substantially parallels Rule 11,” and quoting the Ninth Circuit’s warning in Lnu v. Blanche to “[B]e aware of the risks of overreliance on generative AI.”
The order lists mitigating considerations, including prompt correction, full responsibility, a request that any court sanction fall on him alone, and cooperation with OED, and the resolution was a public reprimand. Firms documenting compliance may wish to consider that practitioners registered before the USPTO answer to that agency’s disciplinary rules for litigation filings in court, not only for prosecution work.
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.
- Verify record citations as well as legal citations. The USPTO notice cautions that AI citation errors may extend to the intrinsic evidence of patent and trademark applications and their file wrappers.
- Review whether a second AI tool is being relied on as the check. Here one AI tool drafted and another reviewed, and the stipulated violations still include not adequately reviewing the citations to the intrinsic record.
- Consider that professional discipline can follow even when the tribunal imposes no sanction. The order notes the district court did not sanction the attorney or his client.