Skip to content 
AI chip designs are transforming physical technology in ways previously unimaginable. Researchers from Princeton University and IIT Madras used AI to generate wireless microchips with unusual yet highly effective layouts—surpassing traditional models in performance. These innovations, however, raise important legal questions: Can AI chip designs be patented? And if no human created the structure, who holds the ownership rights?
For law firms like Stevens Law Group, the question isn’t theoretical. Clients using AI to accelerate design now face unclear rules about intellectual property. This article explores how the law handles AI-created chip designs, what strategies businesses can use, and how legal frameworks might evolve to meet this new reality.
How AI Changes Microchip Design?
Traditionally, building a wireless chip meant connecting electromagnetic elements—like antennas or signal splitters—through manual layout. Engineers would place components one by one and adjust for signal behavior. The Princeton-IIT study disrupted that method.
They trained their AI system to interpret parameters and generate complete microchip layouts. They trained their AI system to generate complete Instead, it created complex structures in minutes that performed better than manually designed chips.
Lead researcher Kaushik Sengupta explained that the designs looked strange—sometimes resembling random shapes—but they worked better than standard chips. These AI-generated chips could operate efficiently across broad frequency ranges and showed energy savings. Unlike humans, the AI viewed the chip as a single object rather than a combination of parts. This approach opened new possibilities in functionality.
These results raise a question: if the machine invents the layout, who can claim the legal rights to it?
Understanding Patent and IP Rules in AI Chip Design
In the U.S., intellectual property protection for chips involves three main tools:
- Patents: These protect functions and structures that are new, useful, and not obvious.
- Copyrights: Limited to visual design or layout under “mask work” protections.
- Trade secrets: Protect confidential methods or designs, if kept secret.
The key requirement for patent law is that an “inventor” must be a natural person. This concept became a legal point of contention in the DABUS case. DABUS is an AI that generated inventions, but when its creator filed patents listing DABUS as the inventor, U.S. and European patent offices rejected them. Courts ruled that only humans can hold inventorship rights.
That precedent now affects all AI-involved innovation. If no human conceived the final design, the current law doesn’t recognize it as protected under patents.
AI and Inventorship: Where the Law Stands
The law today says AI can’t be an inventor. This creates a challenge. If an engineer feeds parameters into an AI and it generates a design, did the engineer invent it—or did the AI? Most likely, the AI created it, which means no human can honestly claim to have conceived the invention.
Still, the USPTO allows humans to be listed as inventors if they made a “significant contribution.” That might mean choosing the AI’s parameters, verifying outputs, or adjusting final results. But if the AI’s design was used exactly as-is, with no creative input from a person, it may not qualify for legal protection.
In the Princeton-IIT case, the researchers confirmed that while the AI handled most of the design, human experts still played a role. They reviewed and corrected the outputs, ensuring they worked as expected. This kind of interaction may be enough to pass the threshold for inventorship under current law.
Why the Patent System Struggles With AI Designs?
AI doesn’t just speed up designs. It makes them different. The Princeton team’s chip structures don’t follow usual patterns. That’s a problem for patents, which require a clear explanation of how something works and why it’s new.
AI designs often defy explanation—even by experts. Without a human-understandable description, it’s hard to meet patent requirements like novelty, non-obviousness, and enablement. If a patent examiner can’t understand the invention, they can’t approve it.
And even if the invention is new, proving that it wasn’t obvious is tricky. What seems obvious to AI may be entirely new to humans, or vice versa. This mismatch makes patent claims harder to support.
Another issue is disclosure. Patents must teach others how to replicate the invention. But if the layout was produced by an AI model using unknown logic, it may be impossible to reproduce without the same system, undermining the application.
Human Oversight: The Legal Workaround
One potential solution is the “human-in-the-loop” method. If a human sets the AI’s constraints, chooses outputs, and checks functionality, they can argue for inventorship. This is the strategy many law firms, including Stevens Law Group, now recommend.
But oversight must be real. Simply pressing “go” on a software tool won’t meet the legal standard. The engineer needs to make decisions that shape the final result.
In the Princeton study, engineers evaluated the AI’s output and discarded configurations that didn’t perform correctly. This sort of oversight may qualify as an inventive contribution.
Companies using AI in chip design should document every human decision in the process. Law firms should help them collect this evidence—from input settings to validation reports—so that patent claims can stand up to legal scrutiny.
Copyright in Chip Layouts
Besides patents, chip layouts can be protected under the Semiconductor Chip Protection Act. This gives 10 years of rights to visual layout designs, known as “mask works.” These protections are limited but useful for companies that want to prevent direct copying.
The problem is that copyright law, like patent law, requires human authorship. The U.S. Copyright Office has recently rejected fully AI-created works. Unless a human makes significant edits to the chip layout, it might not be eligible for copyright.
Again, documentation is crucial. If an engineer tweaks the layout or adjusts the final visual presentation, those changes can justify a copyright claim. But if the AI-generated layout is used without human adjustment, it likely won’t qualify.
Trade Secret Protection for AI-Designed Chips
Trade secrets offer a different kind of protection. There’s no need to register or publish anything. Instead, companies just have to keep the design secret and take steps to protect it, such as using secure systems, access limits, and NDAs.
This method works well for AI-generated chip layouts that are hard to explain. If the AI creates something new, and it performs well, companies may choose to skip patents and protect it as a secret instead.
However, trade secrets don’t stop others from independently developing a similar chip. And if someone leaks the design, the protection disappears. So while this path is useful, it carries risk.
Stevens Law Group and similar firms should advise clients to weigh the pros and cons of trade secrets versus patents. Often, using both makes sense—patenting parts that humans helped design and keeping AI-generated elements private.
What the Princeton-IIT Study Means for the Future?
The study showed that AI can dramatically reduce the cost and time of chip design. Tasks that took weeks now take hours. Designs that once required a team of engineers can now be produced by software.
More importantly, the designs work better. They cover broader frequency ranges and use less energy. The result opens up possibilities for advanced wireless communication, self-driving technology, radar systems, and more.
But this success depends on legal protection. If companies can’t secure IP rights, they may lose the incentive to invest in AI-driven design. Legal teams need to catch up quickly to protect the value AI is creating.
The Path Forward: Updating the Law
Lawmakers and regulators are beginning to explore changes. Some legal scholars have proposed new rules that recognize AI-assisted invention as a separate category. Others suggest redefining “inventor” to include people who use AI as a tool, even if they didn’t directly conceive the invention.
Until then, companies and law firms must work within the current system. That means emphasizing human involvement in the design process, collecting documentation, and building IP strategies that combine patents, copyrights, and trade secrets.
Conclusion
AI-designed chip layouts, such as those developed at Princeton and IIT Madras, are expanding the capabilities of machines. They offer faster development, better performance, and lower costs. But they also expose gaps in current intellectual property law.
For firms like Stevens Law Group, this is both a challenge and an opportunity. Clients need help navigating a system that isn’t fully equipped to handle machine-generated innovation. By guiding businesses on legal strategies, documentation, and filing approaches, IP attorneys can play a key role in protecting the future of chip technology.
FAQs
1. Can AI be an inventor under U.S. patent law?
No. Current rules require that only natural individuals can be listed as inventors.
2. How can companies patent AI-generated chip designs?
They must show that a human played a significant role in creating or validating the design.
3. Are copyright protections available for AI-created layouts?
Only if a human has made creative contributions to the final layout.
4. What if no IP protections apply?
Companies can use trade secret protections, provided they maintain confidentiality.
5. How should law firms adapt?
Advise clients to document all human involvement and consider hybrid IP strategies.
