The Question Behind the Mitchell Order

A few days ago, I wrote about the USPTO's public discipline of U.S. patent attorney Brian E. Mitchell. According to the USPTO OED final order in Proceeding No. D2026-16, one generative AI tool helped draft proposed claim constructions and a second AI tool reviewed the work. The filing still contained numerous erroneous citations to intrinsic patent evidence, including the specification, figures, and prosecution history.

After I shared that article on LinkedIn, an Israeli patent professional asked a simple question: if an examiner makes a similar citation error, does the examiner get disciplined too? He also noted that he was seeing more of these problems across several patent offices. He later sharpened the point: if an applicant has no meaningful avenue to complain, accountability can be difficult to see in practice.

The answer is not a clean “yes” or “no.” At least in the United States, registered patent practitioners and USPTO examiners operate under two different accountability systems.

Two Different Accountability Systems

Registered patent attorneys and agents are regulated practitioners. They must comply with the USPTO Rules of Professional Conduct, and the Office of Enrollment and Discipline can investigate misconduct involving competence, diligence, misrepresentation, and related duties. Serious matters can result in public disciplinary orders, which the USPTO maintains in its OED decisions database.

Patent examiners are different. They are federal employees performing examination duties, not outside practitioners appearing before the Office. If an examiner misreads paragraph [0056] of D1, attributes a feature to a drawing that is not there, or relies on a passage that does not actually support the rejection, that single examination error would not ordinarily produce an OED-style public disciplinary order.

That does not mean examiner errors are consequence-free. The mechanisms are simply different: examination-quality review, supervisory feedback, training, performance management, and—where appropriate—federal employee performance or misconduct procedures.

An Office Action Still Has to Be Grounded in the Actual Reference

The examination rules impose concrete requirements on prior-art rejections. Under 37 C.F.R. § 1.104(c)(2), reproduced in MPEP § 706, an examiner rejecting claims for lack of novelty or obviousness must cite the best references available and, when necessary, identify the particular portion relied upon as nearly as practicable. If the relevance of a reference is not apparent, it must be explained.

For § 103, MPEP § 2142 further expects the Office Action to set out the relevant teachings of the prior art, the differences between the claim and the applied references, the modification needed to reach the claimed subject matter, and the reasoning supporting obviousness.

The search process has a similar source-grounding requirement. MPEP § 904 instructs examiners to study the specification or description of considered references sufficiently to determine the full value of the disclosure relative to the claimed subject matter.

So if an Office Action says D1 discloses a particular technical feature, but opening D1 shows that the feature simply is not there, the first issue is not AI. It is an examination-quality problem.

How the USPTO Reviews Examination Quality

The USPTO has a formal quality-assurance structure for examination work. The Office of Patent Quality Assurance (OPQA) audits a random sample of examination work products each fiscal year for statutory compliance under §§ 101, 102, 103, and 112. The annual review goal is set according to data needs and available resources; USPTO materials in recent years have described goals in the range of roughly 10,000 to 12,000 work products.

Technology Centers and supervisors also conduct quality review, and examination quality is an important part of examiner performance management. The USPTO uses review data to identify recurring issues, trends, and training needs.

In February 2026, the USPTO added another feedback loop. Its monthly Learning from Outcomes to Optimize Patents (LOOP) training series brings together Patents and the PTAB to examine examples of high-quality examination practices as well as potential examination errors identified later in PTAB proceedings. Problems that become visible after examination can therefore feed back into examiner and supervisor training.

So an examiner error does not necessarily disappear into a void. Ordinary work-product problems can enter quality review, feedback, training, and performance-management systems. Persistent unacceptable performance or employee misconduct can implicate separate federal employment procedures, potentially including suspension, demotion, or removal in serious cases.

What Can an Applicant Actually Do When the Examiner Gets the Citation Wrong?

For applicants, this is the more immediate question. If the examiner has plainly misquoted or misread a reference, the first objective is usually to correct the case itself: identify the error in the written response, address it in an examiner interview, and involve the Supervisory Patent Examiner when appropriate.

If normal communication channels have not resolved the problem, the USPTO's Patents Ombuds Office provides another point of contact. The Office describes the Ombuds as a resource for applicants and attorneys when normal processing or prosecution channels have stalled, while making clear that it is not intended to circumvent ordinary communication with examiners, SPEs, or Technology Center management.

The important distinction is that the Ombuds is not an OED-style public disciplinary forum for examiner mistakes, nor does it replace the established mechanisms for deciding the merits of a rejection. If the core dispute is whether the applied reference actually supports the rejection, the applicant still has to work through the prosecution record—response, interview, supervisory involvement, and, after final rejection where appropriate, pre-appeal or formal appeal procedures.

In other words, the United States is not completely without channels for raising examination problems. But those channels are primarily designed to correct the case, restore prosecution to the proper track, and feed quality concerns back into the system. They are not a public individual-accountability process equivalent to OED practitioner discipline.

Three Different Questions: Correction, Quality Oversight, and Personal Discipline

It helps to separate three issues that are often blended together:

  • Case correction: Can the applicant get the erroneous rejection corrected in this application?
  • Quality oversight: Can the error be captured by supervisory review, OPQA sampling, training, or performance-management processes?
  • Personal discipline: Will the public see an individual examiner disciplinary order comparable to an OED order against a practitioner?

The answer to the first two can be yes. The third is a very different question. A practitioner-facing public discipline system and an employee examination-quality system are not the same accountability model.

AI Makes the Difference More Visible

The distinction matters more now because the USPTO itself already uses AI-assisted examination tools.

Beginning in September 2022, utility patent examiners began conducting prior-art searches with the USPTO's AI-assisted Similarity Search (SimSearch) feature. The Office describes SimSearch as an assistive search capability rather than a substitute for examiner judgment, and the examiner remains responsible for deciding whether and how a search result is used.

That is not fundamentally different from a patent lawyer using AI to locate written-description support. If AI tells counsel that paragraph [0065] supports an amendment, counsel still has to open paragraph [0065]. If an AI search tool recommends a reference that looks highly similar to the application, the examiner still has to read the reference and determine what it actually discloses.

At the end of the workflow, someone still has to open the source.

Examiner Error Has a Different Cost for the Applicant

There is an asymmetry in who bears the immediate cost of correction. When a lawyer submits unverified false material to a tribunal, the lawyer may face the client, the court, and professional-responsibility consequences. When an examiner attributes a feature to a reference that does not actually disclose it, the applicant's first task is often to spend additional attorney time and prosecution budget explaining the error in a response or interview.

If that does not work, the applicant may need supervisory involvement, Ombuds assistance where appropriate, and—after final rejection—pre-appeal or formal appeal procedures. From an institutional perspective, those are parts of the patent system's error-correction architecture. From the company's perspective, error correction still consumes real time and legal budget.

One Principle Should Be the Same on Both Sides

The issue is no longer limited to how applicants, patent agents, and lawyers use AI. The USPTO already uses AI-assisted prior-art search tools, and the same verification question will arise wherever a patent office incorporates AI-assisted tools into examination.

The useful principle is therefore broader than any particular professional rule: whoever places AI-assisted content into an official work product should verify the material propositions against the underlying source.

Lawyers, agents, and examiners may face different accountability mechanisms, and applicants may have different correction and complaint routes depending on where the problem occurs. But one point should not change depending on which side of the record the error appears:

AI does not become correct merely because the error has been written into an Office Action.

That leads to a broader question worth continuing to discuss: as AI becomes more deeply embedded in patent examination, should applicants have a more transparent and easier-to-use mechanism not only to correct a particular erroneous Office Action, but also to provide quality feedback and track how recurring examination problems are addressed?

Sources & Further Reading