This is the second installment in a two-part series examining the implications of using generative artificial intelligence (GAI) in the drafting and prosecution of patent applications. In part one, we discussed the privilege and discovery risks that arise when GAI tools are integrated into the patent-drafting process, and we identified targeted discovery strategies for patent litigators challenging patents drafted, at least in part, using GAI tools. In this part, we address the potential impact of drafting a patent application with a GAI tool, including potential invalidity challenges based on anticipation and written description. Thereafter, we offer practical guidance for patent prosecutors and inventors seeking to use GAI in their workflows.
The use of GAI tools as part of the patent-drafting process is not merely an issue of discoverability. It also raises questions regarding patent validity: when GAI generates or substantially contributes to the content of a patent application, can the resulting patent withstand validity challenges? The answers to these questions are not yet settled, and the following discussion outlines both the potential risks and the counterarguments that may be raised.
Initially, and fundamentally, current law requires that a patent list only human inventors who actually conceived the invention. Given the interplay between humans and GAI tools, questions may arise regarding the line between GAI as a tool that assists human conception and GAI as an autonomous contributor. Setting this threshold issue aside, there are other considerations bearing on the validity of a patent drafted with the support of a GAI tool.
First, some commentators have raised the question of whether inputting invention details into a consumer GAI platform could constitute a public disclosure that might affect patentability under 35 U.S.C. § 102. Second, where GAI-generated text populates a patent specification with technical details, embodiments, or examples, questions may arise as to whether the written description demonstrates the inventor's "possession" of the claimed invention as required by 35 U.S.C. § 112(a). Each of these considerations is addressed in more detail below, along with counterarguments that patentees may raise.
Considerations Under Section 102
Under the America Invents Act, a claimed invention is not patentable if it was "described in a printed publication, or in public use, on sale, or otherwise available to the public" before the effective filing date. The touchstone for whether a reference qualifies as prior art is "public accessibility" – whether it was disseminated or otherwise made available such that a person exercising reasonable diligence could have accessed it. A question that has emerged is whether a GAI platform's retention, storage, or use of invention-related inputs could render the invention "otherwise available to the public." This question has not been definitively resolved by the courts.
Consumer GAI platforms, such as ChatGPT and Gemini, may present considerations on this front. Their terms of service may permit the platform to retain user inputs, use them for model training, or handle them in ways that could raise questions about accessibility. Some have argued that if a consumer GAI platform's terms do not guarantee strict confidentiality, inputting detailed invention disclosures could potentially constitute a public disclosure under § 102.
The America Invents Act provides for a one-year grace period before the filing date for any disclosure "made by the inventor or joint inventor." Whether this grace period applies to GAI inputs may depend on the circumstances. If the application is not filed within one year of any potentially qualifying disclosure, the statutory bar would apply. For international filings, practitioners should note that many countries do not offer a similar grace period for public disclosures before the filing date, which may be a consideration for inventors seeking global patent protection. However, whether inputting information into a GAI platform constitutes a "public disclosure" at all remains an open question.
The practical consideration is this: if an inventor or patent prosecutor inputs the substance of an undisclosed invention into a consumer GAI platform before filing, questions may arise regarding whether the information could be subject to discovery or used in validity arguments. However, the success of such arguments would depend on establishing that the disclosure rendered the invention "publicly accessible" under the applicable legal standard – a question that remains unresolved.
Enterprise GAI platforms with contractual confidentiality guarantees and data isolation likely present a materially lower risk on this front. Although an inventor's input into an enterprise GAI platform may still raise discoverability questions, a patent prosecutor's input may be more likely to receive privilege protection, and the confidentiality of the disclosure would likely support an argument that it was not "publicly accessible" or "otherwise available to the public."
Considerations Under Section 112(a)
Another potential invalidity issue arises under 35 U.S.C. § 112(a). Section 112(a) of the Patent Act requires that the specification be written "in such full, clear, concise, and exact terms as to enable any person skilled in the art" to make and use the invention, and that the specification "contain a written description of the invention" sufficient to demonstrate that the applicant was in possession of the claimed invention. The use of GAI in drafting patent specifications has prompted questions about how courts might apply these requirements.
The written-description requirement is distinct from the enablement requirement. While enablement asks whether a person of ordinary skill in the art could make and use the claimed invention based on the specification, the written-description requirement asks whether the specification demonstrates that the inventor actually "possessed" the claimed invention as of the filing date. The test, as articulated by the Federal Circuit in Ariad Pharmaceuticals, Inc. v. Eli Lilly & Co., is an objective inquiry into the "four corners of the specification" to determine whether the disclosure "reasonably conveys to those skilled in the art that the inventor had possession of the claimed subject matter as of the filing date." The level of detail required scales with the complexity and unpredictability of the technology.
This is where GAI-assisted drafting may raise questions. Some commentators have suggested that if a GAI tool generates technical details, embodiments, or examples that the inventor never actually conceived or verified, questions could arise regarding whether the specification demonstrates the inventor's possession of what is claimed. Under this view, a challenger armed with discovery of underlying GAI prompts and outputs might argue that the written description reflects the GAI's synthesis of patterns from its training data rather than the inventor's own understanding.
However, patentees may counter that the written-description requirement focuses on the "four corners of the specification" itself, not on how the specification was drafted. If the specification objectively demonstrates possession of the claimed invention, the tools used in drafting may be irrelevant. Additionally, attorneys have long used various aids – including templates, prior art, technical references, and associate work product – without those aids undermining the validity of the resulting patent. Courts have not yet addressed whether GAI assistance in drafting should be treated differently.
Whether such arguments would succeed in any given case would depend on the specific facts, including the extent of GAI involvement, the nature of the claims, and whether the specification objectively supports the claimed invention.
Practical Considerations for Patent Prosecutors and Agents
Given the evolving questions surrounding GAI-related materials in patent prosecution, practitioners may wish to consider the following approaches. These considerations are not intended to discourage the use of GAI – which offers significant efficiency and quality benefits – nor are they intended to establish any particular standard of practice. Rather, they reflect some of the factors practitioners may wish to weigh in determining how to incorporate GAI tools into their workflows.
- Be Thoughtful About What You Input: Prompts submitted to a GAI platform may be discoverable. Practitioners should advise clients about the risks of utilizing GAI tools in research and development efforts, and practitioners themselves should avoid uploading raw client invention disclosures, attorney mental impressions, or strategic assessments into GAI tools without first considering whether the resulting record could be used against the client in litigation. Where possible, practitioners should use anonymized or abstracted inputs that do not reveal client-specific information.
- Review AI-Generated Outputs: Practitioners may wish to review AI-generated outputs before incorporating them into patent applications or prosecution submissions. This allows practitioners to ensure that filed documents reflect their professional judgment and the inventor's actual knowledge. Where a GAI tool suggests technical details, examples, or claim limitations, confirming with the inventor that those details are accurate may help strengthen the application.
- Consider Documentation Practices: Some practitioners may choose to document the extent of their substantive review and revision of GAI-generated materials. Such documentation, if maintained, may help demonstrate a clear process of independent analysis – receiving the invention disclosure, understanding the technology, consulting with the inventor, and using GAI as one of several drafting aids.
- Evaluate Platform Selection: Given the privilege and discoverability questions discussed in Part 1, practitioners may wish to evaluate the GAI platforms they use. Enterprise GAI platforms that offer data isolation may be more likely to preserve confidentiality. Practitioners who choose to evaluate platforms may wish to review data-processing agreements, paying particular attention to whether inputs are transmitted externally, whether the vendor retains interaction logs, and whether vendor personnel can access user data.
- Be Aware of Potential Discovery Issues: Practitioners should be aware that opposing counsel may request GAI-related materials in litigation. This awareness may inform decisions about when and how to use GAI tools. Practitioners may also wish to discuss with clients the considerations relevant to using consumer GAI platforms for sensitive invention-related work, so that clients can make informed decisions.
Conclusion
The integration of GAI into patent practice is inevitable and, in many respects, beneficial. GAI tools can enhance efficiency and quality in patent prosecution while also raising novel questions that courts and practitioners are still working to address. Both patent prosecutors and litigators should be aware of the evolving legal landscape surrounding privilege, discoverability, and validity as they relate to GAI-assisted patent drafting. As courts continue to address these issues, the legal standards in this area will become clearer. In the meantime, informed awareness of these considerations – rather than avoidance of GAI tools altogether – may be the most practical approach.
For questions regarding the use of GAI, intellectual property strategy, or risk mitigation, please contact Nicole Berkowitz Riccio, Dominic Rota, or any member of Baker Donelson's Intellectual Property Group.