On 11 April 2024, the United States Patent and Trademark Office (USPTO) published guidance (referred to herein as the Guidance) on the use of artificial intelligence (AI) based tools, including generative AI, in practice. The USPTO recognizes the benefits of AI and while practitioners are not presently required to disclose whether AI is used as a drafting tool there are a variety of duties that arise with its use. The Guidance outlines the current USPTO policies and illustrates how these rules interact with the use of AI tools. Below, we will highlight different uses of AI tools and provide an overview of potential risks the USPTO discusses in the Guidance.

Using AI Tools for Drafting or Filing Patent Specifications, Claims, or Office Action Responses

AI tools used in the invention creation process

When an AI tool is used in creating the invention, each claim must have significant inventor contribution. See Inventorship Guidance for AI-Assisted Inventions, 89 FR 10043. If an AI tool is utilized in drafting an application, the AI drafting tool could produce alternative embodiments. In a scenario where a practitioner is uncertain whether the claim has enough contribution by the inventor to amount to “significant,” this should be disclosed to the office. The duty of disclosure requires practitioners to reveal information material to patentability to the USPTO. 37 C.F.R. § 1.56(b). Likewise, practitioners have a duty to not file or prosecute patent claims that are known to be unpatentable.

AI tools used when filing a document or bringing a proceeding

All correspondence to the USPTO must be signed by a natural person. By signing a document, practitioners accept a duty to review and verify the contents of the paper. Relying on the accuracy of an AI tool is unreasonable. 37 C.F.R. § 11.18(b). The USPTO understands that AI tools sometimes “hallucinate” or output incorrect information, so practitioners signing a document must verify the information provided (e.g., citations, technical information, applications disclosed to claim priority, factual assertions, etc.) is true and accurate to their own knowledge. Filing a paper with the knowledge of errors or material admissions violates the duty of candor and good faith. 37 C.F.R. § 11.18.

AI tools used to populate an information disclosure statement (IDS)

AI tools also have the capability to gather prior art references to cite in an IDS. The USPTO warns practitioners that the use of these tools has the potential to increase the size of IDS submissions. Practitioners should never submit an IDS to the USPTO without verifying every reference and removing irrelevant prior art. Submitting an IDS without reviewing each reference cited not only violates a duty under 37 C.F.R. § 11.18, as discussed above, but also the duty not to overburden the examiner. This also applies to third party preissuance submissions.

AI tools used to file documents

The USPTO understands AI tools could be utilized to file documents. The USPTO cautions against using AI in this manner. Not only must signatures be made by a natural person but the USPTO’s policies regarding their electronic filing system (EFS) limit natural persons to holding a USPTO.gov account. Only practitioners with a USPTO.gov account are authorized to file or access documents in the USPTO’s EFS system.

Other Concerns Stemming From the use of AI Tools

Disclosure of confidential information

In using AI tools, practitioners are likely to input confidential information. This confidential information may be retained by the AI tool vendor or may be become part of the AI tools’ training data. In these instances, there is a risk that the confidential data could be disclosed as part of another practitioner’s output. The AI tool vendor could also disclose confidential information to third parties. The USPTO warns any unwarranted disclosure of confidential client information is a violation of a patent practitioner’s duty of confidentiality to their client. 37 C.F.R. § 11.106. When practitioners are considering incorporating AI tools, such as third-party AI patent drafting software, into their practice, they should be particularly cautious about how the vendor can use their data and close attention should be paid to the vendor’s privacy policies, terms of use and/or other contractual terms.

National security concerns

The use of AI tools can also implicate national security or foreign filing license issues where the AI tools are using servers outside the United States. The USPTO’s Manual of Patent Examining Procedure (MPEP) requires applicants to obtain a foreign filing license before “technical data in the form of a patent application, or in any form, are being exported for purposes related to the preparation, filing or possible filing and prosecution of a foreign application.” 37 C.F.R. § 5.11(c). If any AI tool utilizes servers or is hosted by non-U.S. persons, this could be a breach of the MPEP’s foreign filing license requirement. Additionally, “AI system developers or maintainers may suffer data breaches, further subjecting user data to disclosure risks.” Practitioners should understand the terms of use and/or other contractual terms discussing privacy and cybersecurity to mitigate these risks.

The Guidance can be found at Federal Register: Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the United States Patent and Trademark Office.