Can AI file a patent: the legal question and what patent offices have decided

01The question of AI as an inventor

Patent law separates the inventor from the owner. An inventor is the person who contributes to the conception of the claimed invention; an assignee or employer may later own the patent. Generative systems complicate the vocabulary because they can search, combine, optimize, and propose designs without being legal persons.

The central question is not whether AI helped. Engineers have always used tools. The harder question is whether a human made the legally significant contribution to conception, or whether the system generated the inventive concept without a human inventor who can truthfully be named.

02What the USPTO has ruled

The United States Patent and Trademark Office has maintained that current patent statutes require an inventor to be a natural person. The Supreme Court reinforced that reading in Thaler v. Vidal, holding that the term inventor refers to an individual rather than an AI system.

That does not mean AI-assisted inventions are unpatentable. A human who meaningfully conceived the claimed invention can be listed as inventor, even if software helped explore the design space. The application must still satisfy novelty, non-obviousness, enablement, and other ordinary requirements.

03What international patent offices have decided

Patent offices in the United Kingdom, European system, and Australia have likewise rejected attempts to list DABUS as the inventor under their present statutes. The reasoning differs in detail, but the common result is that an AI system cannot be named in the inventorship field.

International divergence remains possible on related questions: how much AI assistance a human can use, what disclosure is expected, and how ownership should be established. Applicants operating globally therefore need an inventorship record that explains the human contribution in every relevant jurisdiction.

04The DABUS case and its implications

DABUS, created by Stephen Thaler, was presented as the inventor of two products in patent applications around the world. The cases became a stress test for language that assumes an inventor is a person. Courts and offices largely answered that the existing framework cannot be expanded by administrative interpretation.

The cases also exposed a distinction between inventorship and disclosure. An AI may be important evidence of how an idea was generated, but naming it does not solve who owns the application, who can sign declarations, or who is accountable for the truth of the filing.

05Why the human inventor requirement exists

Human inventorship connects the patent bargain to a responsible claimant. The inventor can make declarations, assign rights, explain the contribution, and be accountable for material statements. It also gives courts and the public a workable way to resolve priority, ownership, and derivation disputes.

The rule is not simply philosophical. A patent is a legal instrument with enforcement consequences. Treating an unaccountable software process as a rights-bearing inventor would leave basic questions about consent, capacity, ownership, and remedies unanswered.

06What happens when AI creates something patentable

A company should preserve records showing the prompts, designs, experiments, selection decisions, and human revisions that led to the claims. The goal is not to pretend the tool did nothing; it is to identify the people who made the inventive contribution and to avoid an inaccurate inventorship declaration.

Patent counsel can then map the contribution to the claims. If no human can honestly be identified as the inventor under current law, filing may be risky even if the output appears novel. Trade-secret protection, defensive publication, or additional human development may be alternatives depending on the facts.

07How patent law may need to change

As systems become more autonomous, legislatures may revisit inventorship, ownership, disclosure, and incentives. One option is to keep human inventorship but clarify AI-assisted practice. Another is a new legal category with carefully bounded rights and duties. Either approach would require international coordination to avoid incompatible filing games.

For now, the law is clearer than the technology: AI can be a powerful inventive tool, but it is not an inventor in the patent offices that have decided the leading cases. The practical advantage belongs to teams that document human judgment rather than hiding it behind automation.