Artificial intelligence and patent law: what PCB designers and manufacturers need to know

Why this matters now

AI now shows up at almost every stage of making a printed circuit board, and of patenting the results. Engineers use AI-driven tools for component placement and routing, stackup and impedance planning, signal-integrity, power-integrity and thermal simulation, design-for-manufacturing checks, and automated optical and X-ray inspection on the production line. Lawyers and patent examiners use AI to draft, search and review applications. For PCB designers, fabricators and assembly houses, that raises practical questions. Can a board design, fabrication process or inspection method developed with AI be patented? Who is the inventor? Who owns the result?

In the United States, AI-assisted inventions can be patented, but only people can be named as inventors. Patent rights pass from those inventors to their employers or other assignees in the usual way. This article covers how AI is used in PCB development and in patent practice, what the USPTO has said about it, and how inventorship and ownership work when AI is part of designing or building a board.

How AI is used in PCB development and patent practice

AI appears on both sides of the patent system. PCB engineers and manufacturers use it to design, fabricate, assemble and test boards, and patent prosecutors and examiners use it to do their work.

Across all of these, the law still holds people responsible. A person has to contribute to the invention, and a practitioner has to stand behind every paper filed with the USPTO. That holds even when an AI placement tool or inspection model was part of the workflow.

The USPTO’s AI guidance in brief

The USPTO treats AI as a tool. Only natural persons can be inventors, and the same inventorship and patentability rules apply whether or not AI was used, so a board layout refined with an AI tool is examined like any other. The guidance behind that position is summarized below, newest first.

Inventorship

The November 2025 revised guidance says there is no separate or modified standard for AI-assisted inventions. The USPTO presumes that the people named on the application are the actual human inventors. The test is conception: a person must have formed a definite and permanent idea of the complete invention. When several people contribute, the ordinary joint-inventorship rules apply among them. The USPTO no longer compares a person’s contribution with the AI’s. For a PCB team, that means asking who conceived the claimed layout, stackup or process, whatever share of the work the tool did.

Using AI to prepare filings

The April 2024 practitioner guidance does not forbid AI drafting tools and creates no general duty to disclose that one was used. It reminds practitioners that the people who sign a filing must review it, and that relying on an AI tool’s accuracy is not a reasonable inquiry. AI use has to be disclosed if it is material to patentability.

The guidance also raises confidentiality. An AI tool can expose unpublished invention details, such as schematics, stackups, Gerber or ODB++ files and process settings, so its terms of use and data practices matter.

Eligibility of AI-related PCB inventions

With Desjardins and the December 2025 memos, the USPTO has become more receptive to AI and machine-learning claims under § 101. Eligibility is treated as a technical question about how the invention works, and software and AI are not barred as a category. Claims to a physical board, its layer structure or a fabrication process usually do not raise eligibility problems. The harder cases are claims to AI-based design, DFM analysis or inspection methods. Applications that clearly describe the technical improvement, such as fewer false calls in optical inspection or better signal integrity in the routed board, are in the best position.

USPTO guidance governs examination, but it does not bind the courts. A patent granted under today’s guidance can still be challenged in litigation, so good records and careful drafting still matter.

Who owns AI-assisted inventions, and who is the inventor?

Under U.S. law, an AI system cannot be an inventor or a patent owner. Rights start with the human inventors and move to employers or other assignees by contract. If no person qualifies as an inventor, there may be no patent at all.

AI cannot be an inventor

The leading U.S. case is Thaler v. Vidal (Fed. Cir. 2022). Dr. Stephen Thaler filed applications naming his AI system, DABUS, as the sole inventor. The Federal Circuit held that the Patent Act requires inventors to be natural persons, and the Supreme Court declined to hear the case.

Courts in other major jurisdictions have reached the same result. The UK Supreme Court ruled unanimously in December 2023 that an inventor must be a natural person. It noted that the outcome might have been different if Dr. Thaler had named himself as an inventor who used DABUS as a tool. Germany’s Federal Court of Justice held in 2024 that a person must be named, but it allowed a note saying the person had prompted an AI to generate the invention. A 2025 Swiss decision went a step further and treated someone who shaped the AI’s data processing as a co-inventor.

Europe, Australia and New Zealand have also rejected AI inventors. South Africa, which does not substantively examine patent applications, granted a DABUS patent in 2021.

Copyright works the same way. In Thaler v. Perlmutter (D.C. Cir. 2025), the court held that copyright registration requires a human author. That is worth remembering for firmware, documentation and other copyrightable material that AI tools help produce.

When a person using AI is an inventor

Under current USPTO guidance, the question is the familiar one: did a person conceive the claimed invention? Using AI does not disqualify anyone, any more than using an autorouter, a field solver or simulation software would. A PCB engineer is more likely to be an inventor if they:

1. defined the specific technical problem, such as crosstalk on a high-speed interface or warpage in a thin multilayer board, and decided how the AI would be used to solve it;
2. selected, modified or combined AI-generated placements, routes, stackups or process settings into a definite, workable design;
3. recognized which output solved the problem and why, and understood every element of what is claimed; or
4. built, fabricated or tested prototype boards in a way that shaped the final design.

The risk is highest when the AI produced the solution more or less on its own and a person simply ran the tool or accepted the answer, for example by taking an autorouted layout or an AI-suggested reflow profile without change. Owning or supervising the AI system, or typing a generic prompt, is unlikely to be enough. The USPTO’s presumption that the named inventors are correct can be challenged in court, and incorrect inventorship can make a patent invalid or unenforceable.

How ownership works

Once a human inventor is identified, ownership follows the ordinary rules. Patent rights belong first to the named inventors. Employees and contractors usually assign their inventions to the company through employment, consulting or invention-assignment agreements, and those agreements should cover inventions made with AI tools. Contract designers, outside fabricators and EMS providers deserve particular attention: without a written assignment, rights in an improvement conceived by a supplier’s engineers stay on the supplier’s side.

Using a developer’s AI tool, including AI features built into EDA software, does not give the developer ownership of your invention. Vendor terms of service vary, though. Some address who owns outputs, whether your inputs can be used for training, and confidentiality. Read them before uploading design files or other unpublished technical information to any AI system, and that includes AI-enabled DFM checkers and online quoting portals. When several companies contribute data, models or expertise to a project (an OEM, a board fabricator and an inspection-equipment vendor, for instance), agree in advance who will own the resulting inventions.

Output that no person invented cannot be patented. A company can still protect it as a trade secret, which often suits fabrication and assembly process recipes, or publish it defensively.

Practical steps for inventors

A few habits adopted now will make AI-assisted inventions easier to patent and harder to challenge later.

1. Record the human contribution as the work happens. Engineering notebooks, EDA design revisions and version control, prompts, simulation runs and design-review notes should show how people framed problems, judged AI outputs and shaped the final layout or process.
2. Tell your patent counsel how AI was used and which design, simulation or inspection tools were involved.
3. Adopt an internal AI-use policy that says which tools may receive confidential design files and process data. For EDA, DFM and inspection software, review the vendor terms on data use, training and output ownership.
4. File before you share. Putting unpublished design details into a public AI tool, or sending them to a fabricator or customer without a confidentiality agreement, can put confidentiality at risk and, in some cases, novelty.
5. Draft AI-related claims around the technical improvement. Explaining how the invention improves the board, the design tool or the manufacturing process, for example lower signal loss, higher yield or more accurate defect detection, gives the strongest footing under current eligibility guidance.
6. Decide ahead of time what happens to AI output that no one invented, such as an AI-generated layout or process recipe: keep it as a trade secret or publish it defensively.

Where things stand

The USPTO now treats AI as a tool, applies one inventorship standard to every invention, and is more open to claims on AI technology. For PCB designers and manufacturers, the practical message is to use AI freely but make sure a person can show how they conceived what is claimed. The law here is still developing, and the courts may refine these rules.

This article is for general informational purposes only. It is not legal advice and should not be relied on as legal advice for any particular situation. The application of patent law depends on the specific facts, and readers should consult a qualified patent attorney before acting on anything discussed here.

Reading this article, or contacting us about it, does not create an attorney-client relationship. Please do not send confidential information, including details of unfiled inventions, until an attorney-client relationship has been established in writing.

This article reflects the law and USPTO guidance as of October 2026. Patent law, court decisions and agency guidance change, sometimes quickly, and we are under no obligation to update this article after publication. The discussion focuses on U.S. law. References to other countries are general summaries, and the law in each jurisdiction may differ.