Can an AI Be Listed as an Inventor on a Patent?

Current Law Requires a Human Inventor

Patent law, as currently applied in the US and most other jurisdictions, is built around the idea that an inventor must be a natural person — a human being capable of conceiving an invention. This requirement predates modern AI by a long way, but it has been directly tested in recent years as AI systems have taken on a larger role in generating novel ideas, designs, and technical solutions. The US Patent and Trademark Office has addressed the question directly and taken the position that an AI system cannot be listed as an inventor on a patent application, because inventorship under current law is limited to natural persons.

This doesn’t mean inventions developed with significant AI assistance can’t be patented at all. It means the patent needs to identify a human being — typically the person or people who directed the research, interpreted the AI’s output, made key inventive judgment calls, or otherwise contributed genuine inventive input — as the legal inventor, rather than naming the AI system itself.

Why This Question Actually Went to Court

This isn’t a purely theoretical question — it has been tested through real, publicly known cases in which patent applications specifically listed an AI system as the sole inventor, with the human involved arguing the AI system itself had done the genuinely inventive work. Patent offices, including the USPTO, rejected these applications, generally holding that inventorship requires a natural person under existing patent law and statutory language. These decisions have been challenged and reviewed, and the rejections have generally been upheld, reinforcing the position that current law does not permit an AI system to be recognized as an inventor, however sophisticated its contribution.

The reasoning behind this isn’t simply bureaucratic tradition. Patent law ties inventorship to legal concepts like the right to apply for and hold a patent, assign rights, and be held accountable — concepts that are built around legal persons, not software systems. Extending inventorship to an AI system would raise complicated follow-on questions about who owns the resulting patent rights, who can be sued or held liable in patent disputes, and how oaths or declarations required in the patent process would even apply to a non-human entity.

What This Means for AI-Assisted Research and Development

For companies and researchers using AI tools as part of the R&D process, the practical guidance emerging from these cases is that the human contributors involved need to be identifiable and need to have made a genuine inventive contribution, not simply reviewed or rubber-stamped whatever the AI produced. A researcher who uses an AI tool to generate candidate molecular structures, then applies scientific judgment to select, refine, and test the most promising ones, has a stronger claim to inventorship than someone who submits an AI’s raw output essentially unchanged.

This has prompted broader policy conversations about whether patent law will eventually need updating as AI tools become even more central to scientific discovery. For now, though, the clear rule is that patents need a human name attached as the inventor, regardless of how much of the underlying creative work an AI system performed.

Bottom Line

Under current US patent law and guidance, an AI system cannot be listed as an inventor — inventorship requires a natural person — though a human who directs and meaningfully contributes to an AI-assisted invention can still be named as the inventor on the resulting patent.