On April 11, 2024, the U.S. Patent and Trademark Office (USPTO) issued Guidance on Use of Artificial-Intelligence-Based Tools (Guidance), which applies existing rules and policies to the use of artificial intelligence (AI) in practice before the USPTO. The Guidance was issued pursuant to President Biden’s October 2023 “Executive Order on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence,” which calls on government departments and agencies to enact and enforce protections against AI-related harms. Multiple agencies have now issued guidance under the executive order. The USPTO’s new Guidance does not introduce new regulations, but instead discusses risks associated with using AI for drafting documents, filing documents, and interacting with USPTO systems; provides suggestions for mitigating those risks; addresses implications for confidentiality and national security; and reminds applicants and practitioners that the USPTO’s standing duty of candor and good faith (including the duty to disclose known information material to patentability) covers actions taken with AI tools.
 

Drafting Documents for Submission with the USPTO
 

The Guidance provides specific suggestions for reviewing AI-generated content to be included in a USPTO filing. For example, when drafting briefs and office action responses, the Guidance says practitioners should check the accuracy of any citations and ensure that all arguments are legally and technically sound. When using AI to draft patent applications, practitioners should take extra care to verify technical accuracy, compliance with 35 U.S.C. 112(a), and proper distinction between prophetic and working examples in application text. And when using AI to prepare information disclosure statements, practitioners should verify that all cited art is relevant and that no relevant art is omitted.

Practitioners must also avoid submitting AI-generated images and specimens as evidence of use of a trademark when filing new applications, maintaining and renewing trademark registrations, and responding to office actions. And practitioners should use caution when filing AI-generated evidence in proceedings before the U.S. Trademark Trial and Appeal Board. AI-generated images do not show actual use of a trademark in the marketplace and thus are considered misstatements of fact. Further, AI-generated materials submitted as evidence that include irrelevant information or are unnecessarily cumulative may also be thrown out as improper.
 

Using AI to Interact with the USPTO’s Electronic Systems
 

AI systems and tools are not considered “users” for purposes of filing and accessing documents via the USPTO’s electronic filing systems. Thus, AI cannot obtain a USPTO.gov account, cannot have a USPTO Patent Center account, cannot be sponsored as support staff, and cannot sign filings. Furthermore, AI systems should not be used in any manner that generates high numbers of USPTO database accesses or otherwise violates the USPTO’s terms of use.
 

Confidentiality and National Security Concerns
 

According to the Guidance, practitioners must ensure that an AI provider’s terms of use, privacy policies, and cybersecurity practices protect confidential information from being inadvertently disclosed or exported. To do so, practitioners should confirm that AI systems secure data appropriately, do not use client information to train models, and do not send client information to third parties. Practitioners should also obtain clearance before allowing an AI tool to send data abroad, noting that the right granted by a foreign filing license is limited to international filings and does not provide the export clearance an AI tool would need.
 

Duty of Candor and Good Faith
 

The Guidance provides numerous examples of how the USPTO’s duty of candor and good faith applies to AI use. For example, practitioners must ensure their use of AI does not violate their duty to have the legal, scientific, and technical knowledge to represent their clients. Practitioners must also perform their own reasonable inquiry when signing correspondence with the USPTO, as relying on an AI tool does not constitute a reasonable inquiry.

While there is no general duty to inform the USPTO of AI use, the USPTO requires disclosure of any use of AI that is material to patentability. If AI is used to draft claims of a patent application, practitioners must notify the USPTO if any claimed subject matter lacks significant contribution by a human inventor. This could also arise when an AI system assists with drafting the detailed description of a patent application and introduces alternative embodiments that end up in the claims but lack significant human contribution. The implication of a claim lacking significant human contribution is consequential, as “patent protection may [only] be sought for inventions for which a natural person provided a significant contribution to the invention.” Inventorship Guidance for AI-Assisted Inventions, 89 Fed. Reg. 10043 (Feb. 13, 2024); see also Thaler v. Vidal, 43 F.4th 1207, 1213 (Fed. Cir. 2022) (“[O]nly a natural person can be an inventor, so AI cannot be.”).¹ Patent applicants who intend to use AI to support their technical development should consider internal documentation processes to ensure that they not only comply with their duty of disclosure to the USPTO but also are able to rebut any invalidation attempts others may make in the future regarding potential failure to disclose or insufficient human involvement.

The USPTO’s stance on the human inventorship requirement is consistent with the U.S. Copyright Office’s position on human authorship, which has issued similar guidance on the application of AI-generated materials, requiring sufficient human authorship to support a claim. Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16190 (Mar. 16, 2023); see also Thaler v. Perlmutter, Case No. 22-1564 (D. D.C., August 18, 2023) (“Human authorship is a bedrock requirement of copyright.”).