USPTO Rewrites the Rules for AI Patents: What the New Section 101 Guidelines Mean for Global Innovators

In early 2026, the United States Patent and Trademark Office (USPTO) officially enacted new guidelines on patent eligibility for AI inventions under Section 101 of the Patent Act. The scope of these guidelines extends across generative AI, machine learning, and robotics — fundamentally shaking up the U.S. filing strategies of technology companies around the world, including those based in Japan.

1. What Is Section 101? — The Structure of the Wall Blocking AI Patents

Section 101 of the U.S. Patent Act defines patentable subject matter as “any new and useful process, machine, manufacture, or composition of matter.” However, it excludes laws of nature, natural phenomena, and abstract ideas from patent protection. The persistent problem has been that AI algorithms and machine learning models have frequently been classified as “abstract ideas.”

The watershed moment came with the 2014 Supreme Court ruling in Alice Corp. v. CLS Bank International, which established the Alice/Mayo two-step framework: if a claim is directed to an abstract idea, it must add “significantly more” than the abstract idea itself to be eligible. Following this decision, rejection rates for AI and software-related patents surged dramatically. The core of AI inventions — neural network weight adjustment methods, data preprocessing pipelines, and inference engine optimization techniques — were routinely rejected as “abstract ideas.”

According to USPTO statistics, between 2016 and 2023, the Section 101 rejection rate for AI-related applications averaged 1.4 times that of regular software patents. Japanese tech giants including NTT, Fujitsu, and Sony were among those receiving a string of rejection notices.

2. The Architecture of the New Guidelines — What Has Changed

The 2026 USPTO guidelines introduce substantial modifications to how the Alice/Mayo framework is applied in practice. The key changes fall into three categories.

(1) Clarification of “Practical Application” Standards

Under the previous guidelines, an invention directed to an abstract idea could still be eligible if it demonstrated a “practical application” — but the standard for making that determination was vague. The new guidelines enumerate concrete indicators for recognizing practical applications. For instance, an AI invention that explicitly claims to solve a specific technical problem (e.g., improving defect detection accuracy in a specific manufacturing process) is now more readily recognized as demonstrating a practical application.

Critically, AI inventions can now more effectively claim eligibility on the grounds that they “improve the functioning of the computer itself.” A machine learning model compression technique that demonstrably reduces computational resource consumption compared to conventional approaches, for example, can be argued to constitute a technical improvement to computer functionality.

(2) Strengthened Claim Drafting Guidelines

The new guidelines also provide more specific direction on how claims should be written. For AI inventions in particular, rather than simply reciting “a step of analyzing data,” applicants are strongly encouraged to specify “a step of applying a specific algorithmic process to a specific input data format to produce a specific technical effect.” This represents a meaningful departure from prior practice, where many applicants sought broad claims that omitted algorithmic details. Under the new guidelines, explicitly reciting the specific technical mechanism and effect becomes the key to obtaining eligibility.

(3) Handling AI-Generated Inventions

One of the most closely watched aspects of the new guidelines is the treatment of AI-generated inventions. While the USPTO reaffirmed in 2024 that inventors must be human, the new guidelines build on this position by providing concrete criteria for evaluating the human contribution when AI is used as an “invention tool.”

Specifically, even when AI generates multiple candidate solutions and a human selects, adjusts, and validates among them, the human contribution can be recognized as inventorship if it is “non-obvious.” In an era where generative AI-powered R&D is becoming mainstream, this clarification carries immense practical significance.

3. Comparison with OpenAI’s Patent Portfolio — The Limits of “No Patent” Strategy

To understand the impact of the new guidelines, it is instructive to examine the patent holdings of OpenAI, the company most synonymous with generative AI. As of late 2025, OpenAI held only approximately 110 patents — orders of magnitude fewer than Google’s roughly 40,000 and Microsoft’s approximately 30,000.

The reasons are multifaceted. First, OpenAI was founded as an “open research organization” that prioritized sharing knowledge as a public good, choosing technical leadership through published papers and open-source models over exclusionary patent rights. Second, Section 101 presented a formidable barrier. The core technologies underlying GPT-series models — specific variants of transformer architectures, prompt engineering methods, and Reinforcement Learning from Human Feedback (RLHF) — have all faced significant challenges clearing Section 101 examination.

However, since 2025, OpenAI has dramatically accelerated its patent filings. Now that ChatGPT has become a major commercial product and differentiation from competitors is critical, the vulnerability of a “no patent” strategy has become apparent. As Anthropic, Google DeepMind, and Meta’s AI divisions each aggressively build patent portfolios, OpenAI risks being left in a state of intellectual property vulnerability.

The new guidelines represent a genuine opportunity for OpenAI. Transformer architecture variants and specific RLHF implementations that would previously have been rejected may now achieve eligibility under the new “practical application” standard. Many IP professionals predict that OpenAI’s patent portfolio could grow dramatically over the next two to three years.

4. Impact on Japanese Companies — Urgent Need for Filing Strategy Redesign

For Japanese companies, the new guidelines present both opportunities and risks.

Opportunities

Technologies that were previously abandoned for U.S. patenting due to Section 101 barriers may now be eligible for filing under the new guidelines. In particular, AI applications in manufacturing — AI-powered image recognition for product quality inspection, predictive maintenance systems for production lines, and materials informatics for materials design — are areas where Japanese companies have world-class technology. If these can be patented as “practical applications that improve specific manufacturing processes,” it could directly translate into competitive advantages in the U.S. market.

Risks

Adapting to the new guidelines requires specialized expertise and cost in claim drafting. Collaboration with U.S. attorneys and patent agents well-versed in AI technology is essential, yet AI patent specialists remain scarce globally. Moreover, since the new guidelines encourage disclosure of technical details, companies must carefully consider the risk of exposing core algorithms. The “patent vs. trade secret” decision now demands more careful consideration than ever before.

5. Looking Ahead — Are the Guidelines a Stepping Stone to Legislative Reform?

While the USPTO’s new guidelines represent a meaningful step forward, they do not fundamentally resolve the Section 101 problem. The Supreme Court’s Alice decision remains in force, and Federal Circuit precedents continue to apply. Administrative guidelines cannot bind courts.

Industry voices increasingly hope for legislative reform of Section 101, such as the Patent Eligibility Restoration Act. The new guidelines can be viewed as a “bridging measure” that supports legislative efforts underway in Congress. For global innovators, the most desirable outcome would be a legislative amendment that dramatically improves the predictability of Section 101, ensuring that companies with genuine technological contributions can obtain the patent protection they deserve in the world’s largest IP market.

Conclusion

The USPTO’s new guidelines bring practical direction to the patent eligibility analysis for AI inventions while demanding more sophisticated specification and claim design from applicants. As OpenAI’s “openness over patents” approach is being reconsidered, the importance of building AI patent portfolios in the U.S. market — the primary battleground for AI business — is only growing. Technology companies worldwide should seize this inflection point to redesign their AI patent strategies. Deeply understanding the new guidelines and honing claim drafting skills will be the key to winning the next generation of intellectual property competition.