The first two articles in this series addressed attorney-client privilege and the work product doctrine in the context of AI chatbot usage. Those articles focused primarily on litigation and general legal practice. This article turns to a context where the stakes are, in some respects, even higher: patent application drafting and prosecution.
When an inventor or a patent attorney uses an AI tool like Claude®, ChatGPT®, or Gemini® in the course of developing, drafting, or prosecuting a patent application, the privilege questions from the first two articles are still present — but they are joined by a distinct and potentially more consequential set of risks that have nothing to do with evidentiary privilege. Those risks involve patent rights themselves: whether the AI-assisted work could affect inventorship, create prior art, compromise trade secret protection, or trigger disclosure obligations before the United States Patent and Trademark Office. These risks are not hypothetical. They are grounded in current law, and they can be outcome-determinative.
Patent Prosecution Is Not Litigation — and That Difference Matters
The work product doctrine, as codified in Federal Rule of Civil Procedure 26(b)(3), applies in civil litigation. It protects materials prepared by or for an attorney in anticipation of litigation. Patent prosecution is not litigation in that sense — it is a quasi-judicial administrative proceeding before the USPTO with its own rules, its own privilege framework, and its own disclosure obligations.
The application of attorney-client privilege in patent prosecution is well established in broad strokes: communications between a patent attorney or agent and a client for the purpose of obtaining legal advice on patent matters are privileged. But the application of work product doctrine to USPTO proceedings is considerably less settled. The Federal Circuit has addressed patent prosecution privilege in a series of decisions, but questions about the scope of protection for prosecution strategy, prior art analysis, and claim drafting assistance remain actively litigated. Adding AI tools to that already-unsettled landscape introduces new variables that neither the courts nor the USPTO have yet addressed.
There is currently no formal guidance from the USPTO on AI-assisted patent prosecution and privilege. That gap is itself significant — it means that practitioners who use AI tools in prosecution are making judgment calls in a space where the rules have not yet been written.
The Duty of Candor: A Disclosure Obligation That Privilege Cannot Override
Every person associated with the filing and prosecution of a patent application owes a duty of candor and good faith to the USPTO under 37 C.F.R. § 1.56. That duty requires disclosure of information material to patentability — prior art, conflicting claim interpretations, adverse decisions, and similar information that the USPTO would consider important in deciding whether to grant a patent.
The duty of candor creates a tension with privilege that does not exist in litigation contexts, and AI tool usage sharpens that tension considerably. When an attorney or inventor uses an AI chatbot to analyze patentability, search prior art, evaluate claim scope, or compare an invention to existing technology, the AI may surface information that is material to patentability. If that information would be material to patentability under § 1.56, the duty of candor may require its disclosure to the USPTO — regardless of whether the AI-assisted analysis that surfaced it would otherwise be protected as attorney work product or privileged attorney-client communication.
This is not a theoretical concern. AI tools are remarkably effective at identifying prior art, recognizing claim scope issues, and surfacing analogous technologies. The more useful the AI tool is in prosecution, the more likely it is to identify material information. And the duty of candor obligation runs to that information, not to the process by which it was found.
The intersection of AI-generated analysis and the duty of candor is an area where careful thought — and experienced counsel — is essential before AI tools are integrated into prosecution workflows.
Prior Art Risk: When AI Platform Terms Threaten Patent Rights
Entirely separate from privilege, AI chatbot usage in patent practice creates a prior art risk that is specific to patent law and that can affect patent rights whether or not any privilege dispute ever arises.
Under 35 U.S.C. § 102, a patent is not available for an invention that was disclosed to the public before the effective filing date of the patent application. The one-year grace period under § 102(b)(1) provides some protection: an inventor’s own disclosure within one year before filing does not constitute prior art against that inventor. But that grace period has important limitations.
The relevant question for AI tool usage is whether inputting unpublished invention details into a consumer AI platform constitutes a public disclosure. The answer depends on facts that are not yet settled by any court or the USPTO: What are the specific terms governing data use for the platform used? Has the user opted out of model training? Could the information be accessed by third parties — through training use, through a data breach, or through legal process? Is the disclosure sufficiently enabling to constitute prior art?
These questions do not have uniform answers, and the risk profile differs meaningfully depending on which platform is used and under what terms. What is clear is that the risk is real, it is patent-rights-determinative rather than merely evidentiary, and it exists independently of any privilege question. An inventor who loses patent rights because of an inadvertent AI-platform disclosure cannot recover those rights by subsequently establishing that the conversation would have been privileged.
Trade Secret Considerations: A Parallel Risk
For inventions that are not yet the subject of a patent application — or that their owners have decided to protect as trade secrets rather than pursue patent protection — AI tool usage raises trade secret concerns that run parallel to but distinct from the prior art question.
Trade secret protection requires that the information be subject to reasonable measures to maintain its secrecy. Inputting detailed technical information about an unpatented, unpublished invention into a consumer AI platform — under terms that reserve the provider’s right to use that information for model training — raises a genuine question about whether reasonable secrecy measures are being maintained. If a court were later asked to determine whether the information qualified as a trade secret at the relevant time, the terms governing the AI platform used and the steps taken to limit data use would be directly relevant to that inquiry.
This concern applies to inventors, to companies with proprietary processes and formulations, and to anyone who uses AI tools to think through technical problems that they have not yet decided how to protect. The decision to use an AI tool is, in this context, also implicitly a decision about trade secret strategy — whether or not the user recognizes it as such.
Inventorship and AI-Generated Subject Matter
The USPTO has issued guidance making clear that AI systems cannot be listed as inventors on patent applications — inventorship requires human conception of the claimed invention. The Federal Circuit affirmed this position in Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022). But the guidance and the case law address the relatively straightforward question of whether the AI itself can be an inventor. They leave open more nuanced questions that are directly relevant to patent practitioners using AI tools in prosecution.
When an attorney uses an AI tool to generate claim language, propose alternative claim structures, identify claim differentiation strategies, or suggest ways to distinguish prior art, the AI’s contribution to the prosecution strategy may become relevant to questions about inventorship, claim scope, and the written description requirement. If AI-generated claim language departs significantly from what the inventor conceived, questions could arise about whether the resulting claims are supported by the inventor’s disclosure or whether they reflect a form of contribution that the patent system does not currently know how to classify.
These questions are at the frontier of patent law and have not been resolved. They are, however, exactly the kind of questions that sophisticated patent practitioners and their clients should be thinking about as AI tools become more capable and more deeply integrated into the prosecution process.
The Prosecution History: A Permanent Record
One feature of patent prosecution that has no direct analog in litigation privilege analysis is the prosecution history — the public record of communications between the applicant and the USPTO that becomes permanently associated with the issued patent and that can be used in subsequent litigation to interpret claim scope.
Statements made during prosecution — including characterizations of prior art, claim scope arguments, and distinctions drawn to overcome rejections — become part of that public record and can limit the scope of patent protection through the doctrine of prosecution history estoppel. AI-assisted prosecution raises the question of how to manage this dynamic: if AI-generated analysis influences the arguments made during prosecution, and those arguments are later used to narrow claim scope in litigation, the role of AI in generating that strategy becomes a potentially relevant fact.
This is not a reason to avoid AI tools in prosecution — it is a reason to use them with the same care and judgment that experienced patent attorneys apply to every prosecution decision. The prosecution history is always a two-edged sword; AI tools add a new dimension to managing it thoughtfully.
What the USPTO Has — and Has Not — Said
The USPTO has issued guidance on AI and inventorship, and has solicited public comment on AI-related patent policy questions. As of this writing, however, there is no formal USPTO guidance specifically addressing the privilege implications of AI-assisted prosecution, the prior art risk created by AI platform data-use terms, or the duty of candor implications of AI-generated prior art analysis.
That silence is the current state of the law. It means that patent practitioners and their clients are operating in a space where the applicable rules are being developed in real time — through litigation, through bar ethics guidance, and eventually through USPTO rulemaking. The practitioners best positioned to protect their clients are those who understand the relevant risks now, before formal guidance arrives, rather than after a problem has already materialized.
Why This Matters More Than It Might Appear
The risks described in this article share a common feature that distinguishes them from the privilege questions addressed in the companion articles: they are outcome-determinative for patent rights, not just for evidentiary protection.
A privilege ruling can be damaging — it can expose litigation strategy, create adverse inferences, or compromise a client’s negotiating position. But a lost privilege ruling does not, in itself, destroy a patent. An inadvertent prior art disclosure can. A failure to disclose material information to the USPTO can render a patent unenforceable for inequitable conduct. A trade secret that loses its protected status cannot be restored. These are permanent consequences, not litigation disadvantages.
That asymmetry — between the recoverable harm of a lost privilege ruling and the unrecoverable harm of lost patent rights — is the reason that AI tool usage in patent practice deserves more careful analysis than AI tool usage in general litigation, and the reason that guidance from experienced patent counsel is particularly valuable before these tools are integrated into prosecution workflows.
A Developing Area That Rewards Early Attention
The intersection of AI tool usage and patent practice is one of the most consequential and least-analyzed areas of law affecting technology companies, inventors, and patent practitioners today. The legal framework is incomplete, the USPTO has not yet issued comprehensive guidance, and the courts have not yet addressed most of the questions this article raises.
That is precisely why early, informed attention to these issues provides a competitive advantage — for inventors who want to protect their innovations, for companies building patent portfolios, and for practitioners who want to serve their clients well in a rapidly changing landscape.
John Ogilvie of Ogilvie Law Firm advises inventors, technology companies, and other clients on patent prosecution, portfolio strategy, and the legal risks associated with emerging technology, including the privilege, prior art, trade secret, and duty-of-candor implications of AI tool usage in patent practice. If you have questions about how these issues apply to your situation, use the Contact form to get in touch.