In the Matter of Brian E. Mitchell

In the Matter of Brian E. Mitchell, USPTO, Proceeding No. D2026-16, Final Order approved July 27, 2026. Copy of primary document: `~/workspace/goals/ai-sanctions-case-monitoring/files/sources/D-160-in-re-brian-e-mitchell.pdf`.

Mitchell used one generative AI tool to draft the plaintiff's portion of a joint claim-construction chart and a second AI tool to review it, in Magpul Industries Corp. v. Mission First Tactical Group, Inc. et al., No. 2:24-cv-5551 (E.D. Pa.). The AI-generated sections contained inaccurate citations to, and quotations from, the underlying patent's intrinsic evidence (the specification, figures, and file history). Mitchell failed to conduct an adequate review of the AI-generated content before it was submitted to the district court.

37 C.F.R. sections 11.101 (competence), 11.103 (diligence), 11.804(c) (misconduct), and 11.804(d) (conduct prejudicial to the administration of justice). The USPTO issued a public reprimand, with the order published in the Office of Enrollment and Discipline reading room and notice in the Official Gazette, under a settlement agreement with the OED Director. Note: the district court imposed no sanction in the underlying case; the USPTO's independent discipline qualifies as the imposed consequence.

Compare AI-drafted citations against the intrinsic record itself, not against the tool's assurances. Two AI tools checking each other is circular validation; only a human reading the specification and file history closes the loop.

---