AI, Inventorship, and the USPTO’s December Redo

In this article, we return to patent law. Recent changes will affect patent applications for AI-assisted inventions. Of course, we won’t see the impact in courts and arbitration for quite some time. Still, the issues are out there and worth considering.

The December 2025 Revision to AI Inventorship Guidance

On December 1, 2025, the USPTO issued its Revised Inventorship Guidance for AI-Assisted Inventions. This replaces the 2024 guidance and reorients the analysis toward long-standing principles. The official version appears in the Federal Register. [1]

So, What Is USPTO Guidance?

Patent Office guidance is an internal map for USPTO examiners. It helps standardize how the Office applies existing law. But it is not the law itself. Congress enacted the Patent Act, and the courts—primarily the Federal Circuit and, on occasion, the Supreme Court—interpret that Act. When USPTO guidance and judicial interpretation diverge, the courts’ interpretation controls. Still, USPTO guidance shapes day-to-day examination, so inventors must pay attention to it if they hope to get a patent in the first place.

What Happened to the 2024 Pannu-Based Approach?

The 2024 guidance borrowed the Federal Circuit’s joint-inventorship test from Pannu v. Iolab Corp. Under that test, to qualify as a joint inventor: (1) a person must have contributed to the conception of the invention in some significant manner, (2) the contribution must not be insignificant in quality when measured against the full invention, and (3) the person must have done more than merely explain well-known concepts. [2]

That test works for human inventors, but it was not designed to evaluate the role of an AI system. Applying it in that context produced uncertainty. How much human participation was enough? Does selecting or steering an AI tool counted as conception? Does too much AI-generated undermine inventorship altogether?

Returning to First Principles: Conception

The 2025 guidance rescinds the 2024 guidance and takes a new approach. Instead of weighing the significance of human contributions relative to AI outputs, the Office returns inventorship analysis to its traditional core: conception. The essential question is this: Did a natural person conceive the definite and complete idea of the invention as claimed? If so, that person is the inventor. If not, naming the AI system won’t cure the deficiency. AI systems can be incredibly capable. But they are still just tools like “laboratory equipment, computer software, research databases, or any other tool that assists in the inventive process,” according to the USPTO guidance.

Joint Inventorship Remains Human-Only

Joint inventorship principles remain unchanged for multiple human contributors. The Pannu factors continue to govern whether two or more people should be listed as joint inventors. The 2025 guidance simply clarifies that those factors do not apply to machine contributions.

Practical Effects for Inventors

For most inventors, the revised framework is probably good news. A person may use AI tools extensively—whether for modeling, drafting, or simulation—without risking loss of inventorship, so long as the human conceived the invention itself.

But note that documentation is important. Applicants should maintain clear records showing how the human inventor directed, selected, refined, or interpreted AI-generated outputs. These details may matter during prosecution and in later challenges to inventorship.

Courts Will Have the Final Word

Although the USPTO’s guidance offers clarity for examiners, it doesn’t settle the underlying legal issues. Courts ultimately interpret the Patent Act, and their decisions are controlling. In Thaler v. Vidal, the Federal Circuit reaffirmed that inventors must be natural persons. [3] Future cases will address more complex circumstances as AI tools grow more autonomous and capable. For now, the USPTO’s December guidance improves predictability but doesn’t resolve all AI inventorship issues.

Speaking of the Patent Act . . .

Speaking of the Patent Act, you might wonder whether lawmakers are likely to step in and rewrite the Act to address AI. For now, the answer is no. Congress has held hearings on AI and intellectual property, but no bills have been proposed that would change the definition of “inventor” or permit an AI system to be named as one.

The main patent bill receiving attention is the Patent Eligibility Restoration Act (PERA). That bill would broaden eligibility for many AI-related inventions, but it doesn’t touch inventorship. And commentators are divided on its prospects. Some point out that PERA has stalled in multiple Congresses, while supporters argue it is essential for innovation and has bipartisan potential. In short, Congress is paying attention, but it is not close to resolving AI inventorship issues.

What’s Next?

The USPTO’s December guidance offers inventors a clearer framework for addressing AI-assisted inventions. It restores reliance on established doctrines while acknowledging the realities of modern research. But the long-term development of AI inventorship law will continue to unfold in the courts. That is where the ultimate interpretations of the Patent Act happen.

AI and This Article

I used ChatGPT-5.1 to find the USPTO revised guidance and summarize it, and to find information on pending legislation. But don’t worry. I checked it and edited it.

Footnotes

[1] Revised Inventorship Guidance for AI-Assisted Inventions, 90 Fed. Reg. ___ (Dec. 1, 2025).

[2] Pannu v. Iolab Corp., 155 F.3d 1344 (Fed. Cir. 1998).

[3] Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022).