Using AI In Your Inventive Process? Proceed to Patent, But Carefully

1. Inventions Involving AI Can Be Patent-Eligible Under Certain Circumstances

Software-based inventions often encounter patenting obstacles under Section 101 of the Patent Act. These challenges can arise if the claimed invention implicates an algorithm, abstract idea, mental process, natural law, or another category of subject matter that is considered ineligible to be patented.

Inventions involving AI are likely to face similar scrutiny. However, recent Patent Office policy has clarified that these inventions are patentable under certain circumstances.

Recent USPTO Policy Supports Patenting AI Innovations

In September 2025, the incoming Patent Office Director authored a precedential decision on the patentability of an AI innovation, Desjardins, expressing a strong policy stance in favor of protecting these technologies.¹

The decision warned against overbroad Section 101 rejections because categorically excluding AI innovations from patent protection could jeopardize America’s leadership in critical emerging technologies.¹

The reversal in Desjardins rested on at least two features supporting patent eligibility: first, the claims integrated an otherwise abstract idea into a practical application; second, the board credited the claims with improving the functioning of the machine learning model itself, including reduced storage requirements, reduced system complexity, and prevention of catastrophic forgetting.²

Claiming Strategy Impacts Patent Eligibility, Including in Biotech and Healthcare

The patent eligibility of AI innovations will likely be evaluated under the same familiar framework used for software claims. Under that line of authority, eligibility often turns on whether the claims are directed to an improvement in computer functionality rather than an abstract idea.³

Not all AI-assisted inventions will qualify. The Federal Circuit has explained that patents merely applying generic machine learning to new data environments, without disclosing improvements to the models themselves, are patent-ineligible under Section 101.⁴ Likewise, claims focused only on converting, manipulating, and displaying data may fail when they do not provide a technical solution to a technical problem in the prior art.⁵

These same principles apply in life sciences and healthcare. In subject matter eligibility guidance, the USPTO distinguished between eligible and ineligible hypothetical claims to AI-enabled personalized medical treatment, emphasizing that the eligible example recited a specific treatment for the disease.⁶

Appeal decisions also show how the Office may reach different outcomes on similar subject matter depending on the nature of the invention and the claiming approach.

• Ex parte Donovan: claims involving an AI-driven method for treating cancer using imaging data were found patent-eligible because they included additional elements that integrated otherwise ineligible subject matter into a practical application, including specific hardware, treatment output, and defined machine learning operations.⁷
• Ex parte Michalek: claims directed to classifying biomarker data were found ineligible because the analyses were treated as mental processes and did not recite an improvement in computer functionality or a treatment recommendation.⁸
• Ex parte Covit: claims directed to determining medical codes from unformatted data were held patent-eligible because they improved the underlying technology of medical coding.⁹
• Ex parte Reicher: claims to reviewing medical images were held ineligible because they did not provide a technical solution to user error or claim a specific improvement in computer capabilities.¹⁰

Subject-Matter Eligibility Declarations Can Help During Prosecution

To overcome Section 101 rejections, the USPTO has encouraged applicants to submit Subject Matter Eligibility Declarations, or SMEDs.¹¹ These declarations may help demonstrate patent eligibility that is not immediately apparent from the claims alone.

Although SMEDs are entirely optional, the Office has stated they can clarify the record and lead to stronger issued patents.¹² Applicants are encouraged to use them specifically for eligibility issues and not to address unrelated prosecution matters.

2. Inventorship Is Based on Traditional Principles of Human Involvement, Regardless of AI Use

An invention must have an inventor. As more inventions are developed with AI assistance, the Patent Office has continued refining how AI affects inventorship. In guidance issued in December 2025, the Office clarified that inventorship remains grounded in traditional principles applicable to natural persons.¹³

The same legal standard for determining inventorship applies to all inventions, whether or not AI systems were used in the inventive process. There is no separate standard for AI-assisted inventions.¹⁴

Inventorship Depends on Human Conception and Contribution

For a solo inventor, the analysis turns on traditional principles of human conception. The question is whether the natural person conceived the invention under the longstanding conception standard.¹⁵

For multiple human contributors, traditional joint inventorship principles still apply, including the Pannu factors, to determine whether each person qualifies as a joint inventor.¹⁶

Only Natural Persons Can Be Patent Inventors

The December 2025 guidance reiterates that AI cannot be named as an inventor on a patent or patent application. Only natural persons can be inventors.¹⁷

As a result, naming an AI system as an inventor or joint inventor will trigger a rejection during prosecution.¹⁸

AI Systems Are Merely Tools

The USPTO recognizes AI systems as tools used by human inventors. Although AI may generate ideas or provide services, it remains analogous to laboratory equipment, software, databases, or other instruments that assist in the inventive process rather than becoming inventors themselves.¹⁹

The 2024 Inventorship Guidance Is Rescinded

Practitioners should note that the USPTO withdrew its earlier guidance. The older approach applied the Pannu framework to AI-assisted inventions, including those with a solo inventor. The Office later clarified that Pannu applies only to multiple natural persons, not to AI systems, because AI cannot be a joint inventor.²⁰

Takeaways for Companies

Although the governing law in this area may continue to evolve, the recent developments provide a practical roadmap for companies seeking to patent inventions involving AI tools. The strength of these patents will likely depend heavily on prosecution strategy, specification support, and documentation of the inventive process.

Proceed Thoughtfully to Patent AI-Assisted Inventions

The USPTO has shown enthusiasm for patenting AI-related inventions, but claiming strategy will likely shape whether those patents survive examination and later challenges.

• Improve on prior methods: show how the invention improves prior functionality of methods or computer systems, and support those improvements clearly in the specification.
• Demonstrate practical application: integrate AI-related concepts into a concrete technical solution with real-world impact.
• Use specific technological connections: in therapeutic contexts, consider reciting a specific drug or treatment recommendation, along with non-generic hardware or system interactions where appropriate.
• Consider a SMED: after a Section 101 rejection, a well-prepared declaration may help explain why the claims are eligible.

These steps may also pay dividends in litigation. Even though issued claims enjoy a presumption of validity, challengers may later argue that the Patent Office overlooked eligibility problems. A robust specification supporting the arguments above can therefore be valuable both in prosecution and enforcement.

De-Risk AI Involvement in the Inventive Process

Companies may want to promote AI use in development, but AI involvement may later become a target in litigation. Patent challengers will be motivated to argue that patents are invalid for lack of inventorship if they believe the invention was AI-invented rather than human-conceived.

These risks may not be fully avoidable, but they can be mitigated. Companies can monitor the role AI plays in their innovation pipeline and document the inventive process carefully to confirm that inventorship remains grounded in traditional human contribution principles.

For example, a company using AI tools in drug discovery can maintain records showing how human inventors conceived the claimed subject matter and how the AI tools merely assisted that process. This documentation may become critical later if challengers seek discovery into how an invention was developed.

A proactive and thoughtful approach to AI tool use may therefore strengthen patent enforceability and enhance portfolio value over time.

The suggestions above may not be advisable or applicable in all circumstances and do not constitute legal advice. Please contact PatentVest to learn more about patenting strategies for inventions involving AI tools.