The U.S. Patent and Trademark Office (USPTO) is no longer merely observing the generative AI revolution from the sidelines; it is actively weaving the technology into the machinery of American intellectual property. In a detailed briefing on April 29, 2026, senior agency leaders revealed a strategic, two-speed rollout designed to accelerate application processing while attempting to insulate the legal integrity of patents and trademarks from the “hallucinations” and errors common to large language models.
The agency’s approach reveals a stark contrast in philosophy between its two primary arms. While the trademark division is moving with aggressive speed—deploying tools that slash processing times from months to seconds—the patent side is operating under a regime of extreme caution, utilizing narrow pilots and “non-negotiable” guardrails to ensure that human examiners remain the final arbiters of innovation.
For IP stakeholders, from Fortune 500 in-house counsel to solo entrepreneurs, this shift signals a fundamental change in how intellectual property is secured. The USPTO is not just changing its internal software; it is evolving the very standards used to judge whether an invention is “obvious” or whether a trademark is too similar to an existing brand.
A Tale of Two Speeds: Patents vs. Trademarks
The deployment of generative AI within the USPTO is characterized by a cautious crawl in patent operations and a sprint in trademark administration. This disparity reflects the differing stakes: a trademark dispute often centers on consumer confusion, whereas a patent grant can dictate the commercial viability of a billion-dollar pharmaceutical or tech breakthrough.
On the patent side, generative AI is not yet authorized for official examination. Instead, the agency is running a controlled experiment. In the final quarter of fiscal year 2025, a USPTO-approved generative AI tool was deployed exclusively to Supervisory Patent Examiners (SPEs). The goal is to determine if AI can improve the quality of “First Office Actions”—the initial response an applicant receives—without sacrificing the legal defensibility of the patent.
Over the next year, the agency will launch isolated pilot projects focusing on high-friction areas of examination, including:
- Claim Construction: Drafting “broadest reasonable interpretation” analyses to define the scope of an invention.
- Invention Summaries: Generating source-anchored summaries to streamline eligibility analyses under sections 101 and 112.
- Prior Art Mapping: Assembling pinpoint citations to novelty and non-obviousness, reducing the manual labor of claim-to-reference mapping.
- Eligibility Scaffolds: Providing step-by-step frameworks for examiners to determine if an invention is patent-eligible.
Deputy Commissioner for Patents Barry Schindler has been clear that AI will not replace the examiner. “We will only deploy where the tools are demonstrably [helping] patent examiners deliver clearer, faster and more defensible work,” Schindler stated, emphasizing that examiners must remain the authors of record.
Conversely, the trademark division has already embraced the technology. Deputy Commissioner Greg Dotson noted that the organization moved from “doing basically nothing” in AI last fall to launching several live tools. The centerpiece is the Classification Agentic Codification Tool (known as “Class Act”), which automates the assignment of international classes and mark descriptions. According to Dotson, this has reduced certain pendency timelines from “five months into five seconds.”
in early April 2026, the USPTO released a highly anticipated image search tool. This allows users with MyUSPTO accounts to upload a design and instantly scan the corpus of federally registered marks—a capability the agency describes as a “dream come true” for clearance work.
The Shifting Legal Standard: PHOSITA and § 101
Beyond the software, the USPTO is grappling with how AI alters the legal definitions of inventorship and obviousness. Central to This represents the concept of the PHOSITA—the “Person Having Ordinary Skill In The Art.”
Historically, the PHOSITA was a hypothetical human benchmark used to determine if an invention was “obvious.” However, as AI tools become standard in research and development, the USPTO is monitoring whether the “ordinary skill” of a human now includes the ability to use AI. If the benchmark for “ordinary skill” rises, it could become harder to secure patents, as more inventions might be deemed obvious to a PHOSITA equipped with AI.
Policy guidance is also evolving rapidly regarding subject matter eligibility under 35 U.S.C. § 101. Acting Deputy Chief Policy Officer Stefano Karmis highlighted several critical updates that stakeholders must track:
| Policy Milestone | Date | Key Impact |
|---|---|---|
| August Memo | August 2025 | Updated instructions for examiners on AI-involved applications. |
| Ex parte Desjardins | September 2025 | Clarified eligibility for inventions created with AI assistance. |
| Inventorship Guidance | November 2025 | Confirmed the legal standard for inventorship regardless of AI use. |
| Rule 132 Guidance | December 2025 | New pathways to address § 101 rejections via declarations. |
Strategic Implications for Businesses
For entrepreneurs and in-house legal teams, the USPTO’s rollout necessitates a change in prosecution strategy. The agency’s move toward AI-assisted examination means that First Office Actions are likely to be more thorough and better cited. This reduces unnecessary back-and-forth but requires applicants to be more precise and substantive in their initial filings.
There is also a looming requirement for transparency. The USPTO is currently evaluating whether applicants must disclose the specific AI models and training data used during the inventive process. Until formal rules are established, legal experts suggest that companies begin documenting AI involvement in R&D now to avoid future challenges to patent validity.
For those focused on branding, the new trademark tools offer an immediate advantage. The “Class Act” tool allows business owners to align their mark descriptions with the USPTO’s preferred format, potentially eliminating the need for costly attorney interventions in simple filings. The image search tool is similarly vital for “knock-out” searches before investing in a new brand identity.
Disclaimer: This article is for informational purposes only and does not constitute legal advice. Intellectual property law is subject to rapid change; stakeholders should consult with qualified legal counsel regarding specific filings or strategies.
The next major milestone in the rollout is scheduled for July 2026, when examining attorneys in the trademark division are expected to receive access to an internal generative AI tool, provided all governance and audit guardrails are finalized.
How is your organization adapting its IP strategy to account for AI-assisted examination? Share your thoughts in the comments or share this article with your legal team.
