Updated 2026-09-05
from the news desk
Can AI be an inventor? Current patent rules keep inventorship human
European and UK materials say an inventor must be human, and the USPTO says U.S. inventorship analysis stays focused on human contribution. The recurring question is not whether AI can assist, but what patent law does when a machine is…
On the materials cited here, an AI system is not treated as the named inventor. The EPO says only a human can be named under the EPC, the UK Supreme Court materials say DABUS was refused because it was not a person, and the USPTO says U.S. inventorship analysis focuses on human contribution even when AI assists.
The short answer in U.S., UK and Europe
The European Patent Office's inventorship page gives the clearest statement: “Can AI be named as an inventor? Under the EPC, only a human being can be named as an inventor.” The same EPO page says applicants under the EPC must identify a human inventor and explain how they obtained rights from that inventor. The EPO adds that AI systems cannot hold legal rights and therefore cannot be recognised as inventors.
In the U.S., the USPTO's FAQ page does not frame the point in exactly the same sentence, but it points in the same direction by telling applicants and examiners to focus on the human side of the work. The USPTO states: “Regardless of the technology used as a tool in the invention-creation process, the inventorship analysis focuses on the human contribution to the conception of the invention.” The agency also says existing inventorship law applies even when AI assisted in creating the invention.
In the UK, the Supreme Court's case summary for Thaler v Comptroller-General of Patents, Designs and Trademarks says the Patent Office refused DABUS as the inventor because DABUS was not a person for the purposes of the Patents Act 1977. The judgment date listed by the court is 20 December 2023.
What the DABUS cases were actually about
The UK Supreme Court case summary says Stephen L Thaler filed two British patent applications for inventions he said were created by an AI machine known as DABUS, and that he described himself in those applications as DABUS's owner, creator and user. The court's summary also states: “It is not in issue whether DABUS created the inventions autonomously.”
That point matters. On the court's own summary, the dispute was not over whether a machine could make a meaningful technical contribution. One reading is that the real issue was narrower and more legal: whether the statute allowed a machine to be named as the inventor, and whether patent rights could flow from that designation. The same summary records the refusal this way: “On 4 December 2019, the respondent handed down a decision refusing to accept the designations of DABUS as the inventor in the Applications pursuant to section 13(2) of the 1977 Act on the ground that DABUS is not a person, as envisaged by sections 7 and 13 of the 1977 Act.”
Europe framed the issue in similar terms. The EPO says its Legal Board of Appeal confirmed the human-only rule in decision J 8/20 after applications listed DABUS as inventor. The EPO also says the EPC requires identification of a human inventor and an explanation of how the applicant got rights from that inventor.
In the U.S., the Court of Appeals for the Federal Circuit page for Thaler v. Vidal states that the court posted a precedential opinion on 5 August 2022. The point that can safely be taken from that page is limited but relevant: the DABUS dispute also reached a precedential U.S. appellate decision.
What still counts as AI-assisted invention
The key U.S. distinction is between AI as a named inventor and AI as a tool used by a human inventor. The USPTO says its guidance does not raise the inventorship bar for AI-assisted inventions. The agency states that the guidance does not create a heightened standard, and it explains that patents function to incentivize and reward human ingenuity.
The USPTO also says existing inventorship law applies regardless of whether a person used AI to assist in creating the invention. In plain English, the presence of AI does not by itself defeat patenting if the legally relevant human contribution is still there. That is why the USPTO keeps returning to human contribution rather than to the mere fact that AI was involved.
The same FAQ page says applicants do not take on a new AI-specific disclosure burden. The USPTO states that the guidance does not impose any additional duty to disclose information beyond the agency's existing rules and policies.
The FAQ page also says examiners ordinarily do not investigate inventorship unless there is some reason to do so. The USPTO notes that examiners typically do not make inquiries regarding inventorship. It adds, citing the Manual of Patent Examining Procedure, that the agency generally presumes the named inventor or joint inventors are the actual inventors, and that rejections for improper inventorship under 35 U.S.C. sections 101 and 115 are rare because of that presumption.
One reading of the USPTO material is practical as well as legal. If a human can properly be named because that person made the human contribution the law requires, the application proceeds under the same basic inventorship framework as other applications. The AI's role may be technically important, but the naming question still turns on the human side of the inventive act.
Why the rule is about law, not just technology
The strongest case for the human-inventor rule, on these materials, is structural. The EPO says the EPC requires a human inventor and an explanation of how the applicant obtained rights from that inventor. The EPO then gives a reason: AI systems cannot hold legal rights. The USPTO gives a related policy explanation when it says inventorship analysis focuses on human contribution because patents function to reward human ingenuity.
On that reading, patent law is not simply asking who or what generated an idea. It is also asking who can be named in a legal document, whose rights can transfer, and who fits inside a statute written around inventors as legal actors. For a related status question, see our legal personhood glossary.
The strongest counter-reading, still grounded in these documents, is that the DABUS litigation exposes a pressure point. The UK Supreme Court's summary says autonomous creation was not in issue. A reader could infer that the harder policy question is what patent law should do if a system is said to have created an invention but cannot be named. The materials cited here do not resolve that policy argument in favor of changing the rule. They do, however, show why the question keeps returning.
The EPO also says this human-only position reflects an international consensus and adds that South Africa is, to date, the only country to have accepted the designation of an AI system as an inventor. As the EPO presents the comparison, the disagreement is not about whether AI can be technically useful. It is about whether a patent system treats an AI system as a rights-bearing inventor. For neighboring doctrines built around human legal status, see our explainers on AI copyright and AI and contract signing.
Frequently asked questions
If I used AI as a tool, can I still be the inventor?
Yes, if the human contribution that patent law requires is still present. The USPTO says existing inventorship law applies regardless of whether AI assisted in the creation of an invention, and it says the analysis should focus on the human contribution.
Do I have to tell the USPTO how much AI I used?
Not under any new AI-specific rule. The USPTO says its guidance does not impose any additional duty to disclose information beyond the agency's existing rules and policies.
Do patent examiners investigate whether AI was involved?
The USPTO says they typically do not make inquiries regarding inventorship. The agency also says it generally presumes the named inventor or joint inventors are the actual inventors, and that rejections for improper inventorship are rare because of that presumption.
Has any country accepted an AI system as an inventor?
The EPO says that, to date, South Africa is the only country to have accepted the designation of an AI system as an inventor. The same EPO page says that under the EPC only a human being can be named as an inventor.
caveats: The EPO page is the sole source used here for the international-consensus point and for the statement that South Africa is the only country to have accepted an AI inventor designation to date. The Federal Circuit point is limited to what the court's case page states: that a precedential opinion in Thaler v. Vidal was posted on 5 August 2022.
Sources
- USPTO — FAQs on Inventorship Guidance for AI-assisted Inventions — 2025-01-16
- U.S. Court of Appeals for the Federal Circuit — THALER v. VIDAL [OPINION], Precedential — 2022-08-05
- European Patent Office — Inventorship — 2026-08-14
- UK Supreme Court — Thaler (Appellant) v Comptroller-General of Patents, Designs and Trademarks (Respondent) — 2023-12-20