Skip to content

Incident Tracker / United States (federal)

In re Brian E. Mitchell, USPTO Proceeding No. D2026-16

·US Patent and Trademark Office, Office of Enrollment and Discipline·US-FEDOther

The US Patent and Trademark Office's Office of Enrollment and Discipline publicly reprimanded California patent attorney Brian Mitchell, whose AI-drafted claim-construction chart in a federal patent case cited nonexistent portions of the patent's intrinsic record. Reported as the USPTO's first disciplinary order predicated on generative-AI misuse, it treats hallucinated citations to a specification and file history as an ethics violation.

Court
US Patent and Trademark Office, Office of Enrollment and Discipline
Date
2026-07-27
Who used the AI
Lawyer
AI tool
Not identified in the record
Conduct
A registered patent attorney (Reg. No. 50,612, California) used two generative AI tools, one to draft and one to review, to prepare a joint claim-construction chart in Magpul Industries Corp. v. Mission First Tactical Group (E.D. Pa.) whose citations to the patent's specification, figures and prosecution history referred to nonexistent or misattributed passages.
Outcome
Other
Penalty
No monetary penalty recorded
Regulatory outcome
USPTO Office of Enrollment and Discipline: Public reprimand by settlement for violations of 37 CFR 11.101, 11.103, 11.804(c) and 11.804(d); settlement executed 20 July 2026, approved 21 July 2026, final order dated 27 July 2026, published on the OED reading room 27 August 2026. (2026-07-27)
Status
Verified against listed sources
Last checked
2026-09-03

What was fabricated or misused

  • Joint Claim Construction Chart citations, quotations and parentheticals to the patent specification, figures and prosecution history referred to portions of the intrinsic record that did not exist or were inaccurately attributed
  • AI-generated citations inaccurately cited to and quoted from the patent’s intrinsic evidence; the final order did not enumerate each citation

What the court said

“Respondent used one generative artificial intelligence (“AI”) tool to assist with drafting proposed claim constructions of certain claim terms of the patent in question. Respondent then used a second AI tool to review the plaintiff's portion of the document.”
USPTO OED final order, Final Order ¶6
“Specifically, many of the plaintiff's citations, quotations, and parentheticals referred to portions of the record that did not exist or were inaccurately attributed.”
USPTO OED final order, Final Order ¶8
“Respondent acknowledged that the erroneous citations were not sufficiently checked against the source documents.”
USPTO OED final order, Final Order ¶10
“Respondent is hereby publicly reprimanded;”
USPTO OED final order, Final Order ¶20(a)

Timeline

  1. Mitchell executed the proposed settlement agreement with OED
  2. OED Director executed the proposed settlement agreement
  3. USPTO approved the settlement and ordered a public reprimand
  4. OED published the final order in its public reading room

Why this case matters

Mitchell is the tracker’s first USPTO discipline row and differs from Mata and LeDoux because the false material was not case law. It was intrinsic patent evidence in a claim-construction chart. The order shows that AI verification duties cover the specification, figures, prosecution history and file wrapper, not only statutes and decisions. It also records mitigation: the lawyer corrected the chart quickly, accepted responsibility and the court imposed no sanction, yet OED still used a public reprimand because the filing had been signed and submitted to a tribunal.

Practice note

For patent filings, verify every AI-assisted citation against the intrinsic record before signing: open the patent, figures, prosecution history and file wrapper, then confirm the quoted text and pinpoint. A second AI tool is not a source check. If errors surface, correct the chart immediately, tell the tribunal and client, and preserve the review trail for OED.

Primary sources for this incident

Related regulation records for this incident

No related regulation record has been linked to this incident yet.

Similar incidents on the record

Cite this record

SafeLegalAI Global Legal AI Incident Tracker, “In re Brian E. Mitchell, USPTO Proceeding No. D2026-16”, safelegalai.com/tracker/in-re-brian-mitchell-uspto (accessed 2026-10-07). Data: CC BY 4.0.

Report an issue with this record →