AI Patent Inventorship Disputes 2026: Can an AI Be Named an Inventor?
The DABUS litigation settled the headline question — AI can't be an inventor. The harder, ongoing question is what counts as "significant human contribution" when AI does most of the inventive heavy lifting, and what happens to a patent that gets the answer wrong.
The DABUS Case Set the Baseline Rule
Inventor Stephen Thaler filed patent applications naming his AI system, DABUS, as the sole inventor of two inventions it allegedly generated autonomously. The USPTO rejected the applications for failing to name a valid inventor, and the Federal Circuit affirmed in Thaler v. Vidal, holding that the Patent Act's use of "individual" for inventor status means a natural person — full stop. The Supreme Court declined to review the case, leaving the Federal Circuit's ruling as controlling law.
Similar rulings followed in the UK, EU, and Australia after early conflicting decisions, converging on the same result: no major patent system currently allows an AI system to be named as an inventor, regardless of how much of the inventive work it performed.
The Question That Actually Matters for Business: How Much Human Contribution Is Enough?
DABUS answered a narrow question — AI alone can't be the inventor. It didn't answer the question businesses actually face: when AI substantially assists an invention, how much did the human involved need to do to be a legitimate named inventor? The USPTO's inventorship guidance answers this with a framework borrowed from joint-inventorship case law (the Pannu factors), applied to human-AI collaboration.
Insufficient: Merely Posing a Problem
NOT ENOUGHSimply presenting a problem to an AI system and having it generate a solution does not make the person who posed the problem an inventor — even if the problem statement was sophisticated and required deep domain expertise to formulate.
Insufficient: Merely Recognizing a Good Output
NOT ENOUGHReviewing AI-generated outputs and picking the one that works best does not, by itself, constitute a significant contribution to conception — recognizing that a solution works is not the same as inventing it.
Sufficient: Constructing the Prompt as Part of Conception
CAN COUNTWhere a person designs a prompt reflecting a specific solution they already conceived, in a way that shows the person possessed the invention before the AI produced its output, this can constitute a significant contribution.
Sufficient: Modifying, Combining, or Reducing AI Output to Practice
CAN COUNTA person who takes AI-generated output and significantly modifies it, combines it with other elements to form the invention, or does the work of reducing an abstract AI output to a practical, working embodiment has likely made a significant contribution.
Why Getting Inventorship Wrong Can Invalidate the Whole Patent
Inventorship isn't a formality — it's a substantive requirement. A patent that names a human inventor who did not, in fact, make a significant inventive contribution is subject to challenge for improper inventorship. If a court or the USPTO finds the naming was done deliberately to obscure that AI (or another actual contributor) did the inventive work, it can support a finding of inequitable conduct, which can render the entire patent unenforceable — not just the improperly claimed portions.
This creates a real business risk pattern: an R&D team uses an AI tool that does most of the conceptual work, an engineer lightly reviews and files under their own name without documenting a genuine contribution, and years later — often during litigation against a competitor infringing the patent — the defendant's counsel challenges inventorship as a way to invalidate the patent entirely.
Documentation Practices That Protect AI-Assisted Patents
Contemporaneous invention disclosure records
Document, at the time of invention, exactly what the human contributed — the conceptual framing, the modifications made to AI output, and the reduction to practice. Retroactive justification is far weaker evidence than contemporaneous notes.
Separate 'AI-assisted' from 'AI-generated' in your R&D process
Build an internal process that distinguishes AI-assisted invention (human conceives, AI helps execute or explore) from cases where AI output is taken essentially as-is — the latter carries much higher inventorship risk.
Train inventors on the Pannu-derived standard
Engineers and scientists using AI tools daily should understand that posing a good prompt or picking a good output is not, by itself, inventorship — and know what documentation actually establishes a significant contribution.
Audit before litigation, not during it
If you hold patents built with heavy AI assistance, review the invention record now. Discovering a documentation gap during an infringement suit, when the patent's validity is under attack, is the worst time to find out.
Frequently Asked Questions
Is this the same rule as AI copyright ownership?
No — patent and copyright law treat AI-assisted creation differently in some respects, but they share the same core requirement of significant human authorship or inventive contribution. Both the Copyright Office and USPTO have converged on frameworks that deny protection to purely AI-generated output while protecting works or inventions with genuine human creative or inventive input.
Can I name myself as the sole inventor if AI did most of the work, as long as I did something?
Only if your contribution meets the significant-contribution standard, and only listing inventors who actually contributed. If multiple people made significant contributions, all must be named as joint inventors; naming only one person who did the least significant human work while omitting others creates its own inventorship problem.
Does this affect patent applications outside the US?
The specific legal test varies by jurisdiction, but the UK, EU, and Australia have reached similar conclusions after their own DABUS litigation — no jurisdiction currently permits an AI system to be a named inventor. Businesses filing internationally should expect the same core documentation practices to matter everywhere.
What if my company doesn't currently have any AI-assisted patents — is this still relevant?
If your R&D team uses AI coding, chemistry, drug-discovery, or design tools at all, you likely have AI-assisted inventions in the pipeline already. Building the documentation habit before your next invention disclosure is far cheaper than reconstructing a defensible inventorship record after a patent is challenged.
Find AI R&D and Documentation Tools on RatedWithAI
RatedWithAI reviews AI-assisted research, coding, and lab-documentation tools — including which ones support the kind of contemporaneous invention records that hold up under an inventorship challenge.
Explore AI Legal & IP Guides