U.S. Patent Rules Face New Scrutiny as AI Inventions and Software Claims Challenge Established Legal Tests

Artificial intelligence is changing how companies develop software, design technical systems and discover new solutions. But as innovation accelerates, a difficult question remains: when does a computer-based invention qualify for patent protection, and who can legally claim to have invented it?

In the United States, patent examiners and courts are confronting several connected challenges. The U.S. Patent and Trademark Office (USPTO) has updated its approach to assessing AI-assisted inventorship and software-related patent eligibility, while court decisions continue to shape the boundaries of what can be patented. Separately, disputes over standard-essential patents are testing how technology companies negotiate licences for inventions required by technical standards.

These developments do not represent a single change in patent law. Instead, they illustrate how established intellectual property rules are being applied to technologies that increasingly depend on software, machine learning and interconnected devices.

Why Software and AI Patents Are Difficult to Assess

Patent law is designed to protect qualifying inventions, not every idea that can be expressed through a computer program.

Under Section 101 of the U.S. Patent Act, an invention generally must fall within a recognised category, such as a process, machine, manufacture or composition of matter. Even if a claim fits one of these categories, it may still be ineligible if it is directed to an abstract idea without sufficient additional elements to establish a patent-eligible application.

This distinction has become especially important for software and artificial intelligence.

A company might develop a new system for analysing medical images, optimising a logistics network or processing financial transactions. The challenge is determining whether its patent claims describe a specific technological improvement or simply apply a mathematical method, business process or abstract concept using conventional computing equipment.

The difference can determine whether a patent application proceeds, faces rejection or becomes the subject of a court dispute.

For technology businesses, the outcome affects more than legal paperwork. Patent protection can influence investment decisions, licensing negotiations, research strategy and the commercial value of software products.

USPTO Updates Its Approach to Software and AI Patent Eligibility

The USPTO has continued refining how examiners evaluate whether computer-implemented inventions qualify for patent protection.

In December 2025, the office announced changes to the Manual of Patent Examining Procedure following the Patent Trial and Appeal Board’s precedential decision in Ex parte Desjardins. The changes address how examiners should evaluate claims that assert improvements to computer functionality, data structures, learning models or other technological fields.

The guidance emphasises that claims should be assessed as a whole, including the technological improvements described in the patent specification. The analysis must still follow the established legal framework for patent eligibility, rather than treating the use of AI or software as sufficient by itself.

The USPTO also issued a September 29, 2026, memorandum outlining a flowchart and best practices for subject-matter eligibility declarations. These declarations allow patent applicants to submit evidence relevant to eligibility under the applicable rules.

The developments provide additional direction for examiners and applicants, but they do not create an automatic right to patent software or AI systems. A claim must still satisfy the relevant legal requirements, and eligibility is separate from questions of novelty, non-obviousness and adequate disclosure.

Courts Continue to Draw the Line Between Software Innovation and Abstract Ideas

Judicial decisions remain central to defining the limits of software patent protection.

In April 2025, the U.S. Court of Appeals for the Federal Circuit ruled in Recentive Analytics, Inc. v. Fox Corp. that the patents at issue, which applied machine-learning techniques to optimise network maps and event schedules, were ineligible under Section 101.

The court found that applying conventional machine-learning techniques to a new data environment did not, by itself, establish a patent-eligible technological improvement.

The decision illustrates a recurring issue for AI developers. A new application of an existing model or algorithm may be commercially useful, but usefulness alone does not necessarily establish patent eligibility.

A stronger patent claim may need to identify a particular technical mechanism or improvement, rather than merely describing a desired outcome and instructing a computer to produce it.

That does not mean machine-learning inventions are categorically excluded from patent protection. The relevant question depends on the specific claims, the technology described and the applicable legal test.

For developers, the practical challenge is to explain what the invention changes technically, how it achieves that change and why the claimed implementation is more than the routine use of computing tools.

Who Is the Inventor When AI Helps Create an Invention?

Patent eligibility is only one part of the challenge. AI also raises questions about inventorship: which person or people can legally be named as the creators of a patented invention?

In November 2025, the USPTO rescinded its February 2024 guidance on AI-assisted inventorship and issued revised guidance. The updated position reaffirmed the longstanding rule that only natural persons can be named as inventors under U.S. patent law.

The office also clarified that the same legal standard for determining inventorship applies whether or not AI tools were used during development.

This means an AI system cannot be named as an inventor or joint inventor. However, the use of AI does not automatically make an invention unpatentable. A human may still qualify as an inventor if their contribution meets the applicable legal standard.

The central issue is the human contribution to the conception of the claimed invention. Simply identifying a general problem and asking an AI system to solve it may not be sufficient. The analysis depends on what the person actually contributed to the specific invention.

This distinction matters as AI tools become more capable of generating designs, suggesting technical solutions and producing code. Companies may need to maintain clear records of how inventors developed an idea, what the AI system contributed and how human decisions shaped the resulting invention.

The revised guidance does not create a special patent category for AI-generated inventions. It applies established inventorship principles to a changing technological environment.

Why Patent Eligibility and Inventorship Must Be Treated Separately

The two questions are related but legally distinct.

Patent eligibility asks whether the claimed subject matter qualifies for protection under the relevant statutory framework. Inventorship asks which human beings legally contributed to the conception of the claimed invention.

An AI-assisted invention could satisfy the technical requirements for patent eligibility but still face an inventorship problem if no named human qualifies as an inventor. Conversely, a human may clearly have conceived an invention, yet the patent claims could still be rejected because they cover an abstract idea or fail another statutory requirement.

This distinction is important for companies using AI in research and product development.

Patent strategy cannot rely solely on documenting that a system produced a useful output. Applicants must also consider how the invention is defined in the claims, whether the description supports those claims and whether the named inventors made the contributions required by law.

The outcome may affect ownership disputes, licensing agreements and the enforceability of patents that underpin commercial products.

Standard-Essential Patents Add Another Layer of Complexity

A separate patent-law challenge concerns standard-essential patents, commonly known as SEPs.

These patents cover technologies considered essential to implementing a technical standard. Such standards enable products from different manufacturers to communicate and operate together. They are important in areas including cellular connectivity, Wi-Fi, connected vehicles and other networked systems.

Because manufacturers may need to use standardised technologies to make compatible products, SEP licensing can become commercially significant.

Many SEP holders commit to licensing their essential patents on fair, reasonable and non-discriminatory terms, known as FRAND. Disagreements can arise over what constitutes a fair royalty, whether an offered licence is reasonable and how a licence should cover products sold across multiple countries.

These disputes can involve complex portfolios, competing valuation methods and questions about the appropriate balance between rewarding patent owners and enabling widespread adoption of technical standards.

SEP licensing is not the same legal question as software patent eligibility. One concerns the terms on which essential patented technologies are licensed; the other concerns whether particular claims qualify for patent protection. However, both influence the conditions under which technology companies can develop and commercialise products.

A Global Dispute With Implications for Connected Vehicles

A significant recent illustration came from the United Kingdom.

In July 2026, the UK Supreme Court issued its judgment in Tesla, Inc. v. InterDigital Patent Holdings, Inc., a dispute involving a global licensing platform for 5G standard-essential patents used in connected vehicles.

The case examined issues surrounding the terms of a global licence, including FRAND obligations and the role of licensing platforms in providing access to large patent portfolios.

The judgment is relevant to companies developing connected vehicles and other products that depend on communications standards. It also highlights how SEP disputes can extend beyond individual patents to the structure of licensing arrangements across international markets.

However, this was a UK Supreme Court decision, not a ruling by the U.S. Supreme Court. Its implications should be understood within the applicable legal framework and distinguished from U.S. patent-eligibility and inventorship disputes.

For multinational technology businesses, the broader lesson is that patent licensing strategy increasingly requires an understanding of different national courts, contractual commitments and cross-border commercial arrangements.

What Technology Companies Should Watch Next

The direction of patent law will matter to companies investing in AI, software, connected devices and communications infrastructure.

For AI developers, the immediate priorities include defining genuine technical improvements, preparing patent claims that explain how an invention works and maintaining evidence of human contributions to inventorship.

For patent applicants, updated USPTO guidance offers more detail on how eligibility questions may be assessed. It does not remove the need to establish that claims satisfy the law.

For manufacturers relying on standardised technologies, SEP licensing remains a separate commercial and legal concern. Companies must consider whether their products use patented standards, how licensing commitments apply and how disputes over royalty terms may affect market access.

The broader trend is not that software and AI patents are becoming universally easier or harder to obtain. Rather, courts and patent offices are refining how established principles apply to new technical circumstances.

As AI becomes more deeply embedded in invention and product development, the distinction between a promising computational result, a patent-eligible technical solution and a legally attributable human invention will remain central to intellectual property strategy.

Frequently Asked Questions

1. Can software and AI inventions be patented in the United States?

Yes, some software and AI inventions can receive patent protection. They must satisfy applicable requirements, including subject-matter eligibility, novelty, non-obviousness and adequate disclosure. Using AI or a computer does not automatically make a claim patentable.

2. Can an AI system be named as an inventor on a U.S. patent?

No. The USPTO’s revised November 2025 guidance reaffirms that only natural persons can be named as inventors. A human who uses AI may qualify if their contribution meets the legal standard for inventorship.

3. What did the Recentive Analytics v. Fox Corp. decision establish?

In April 2025, the Federal Circuit found the patents at issue ineligible because applying conventional machine-learning techniques to a new data environment did not establish the required technological improvement. The decision does not categorically exclude all AI-related inventions from patent protection.

4. What are standard-essential patents?

Standard-essential patents cover technologies required to implement a technical standard, such as certain cellular or Wi-Fi technologies. Licensing disputes often concern royalty levels and the fair, reasonable and non-discriminatory terms on which these patents should be licensed.

5. Does updated USPTO guidance guarantee approval of software or AI patent applications?

No. Guidance helps examiners and applicants apply existing legal standards, but each application is assessed on its claims and supporting disclosure. Patent eligibility is also distinct from novelty, non-obviousness, inventorship and other legal requirements.