AI Inventorship: Can AI Be a Patent Inventor? [2026 Update]

By Michael Meyer, USPTO-Registered Patent Attorney | Chemistry Degree, University of Nebraska Omaha | J.D., Creighton University | Updated March 2026

Artificial intelligence systems are now generating novel chemical compounds, designing mechanical components, creating algorithms, and solving engineering problems with minimal human input. As AI becomes more autonomous and creative, a fundamental question has emerged: Can an AI system be named as an inventor on a patent application? And if not, who owns the rights to inventions conceived primarily by AI?

This is one of the most rapidly evolving areas of intellectual property law. Recent court cases, USPTO guidance, and international patent office decisions are shaping the rules for AI-generated inventions — with significant implications for AI companies, researchers, and anyone using machine learning tools in the innovation process.

This guide explains the current legal status of AI inventorship in the United States, the landmark DABUS case that tested whether AI can be an inventor, what the USPTO requires for AI-assisted inventions, practical strategies for filing patents when AI contributes to the invention, and how international jurisdictions are handling these issues.

Table of Contents

• The Core Question: Can AI Be a Patent Inventor?
• The DABUS Case: AI Inventorship Rejected Worldwide
• Current U.S. Law: Only Natural Persons Can Be Inventors
• USPTO Guidance on AI-Assisted Inventions (2024)
• What Constitutes "Significant Human Contribution"?
• Practical Implications for AI Companies and Researchers
• International Approaches to AI Inventorship
• Future Outlook: Proposed Legislation and Policy Debates
• How to Handle Patent Applications for AI-Generated Inventions
• Frequently Asked Questions

The Core Question: Can AI Be a Patent Inventor?

Under current U.S. patent law, the answer is no — an AI system cannot be named as an inventor on a patent application. The Patent Act requires that inventors be "natural persons" (human beings), and both the USPTO and federal courts have consistently rejected attempts to list AI systems as inventors.

Why This Question Matters

As AI systems become more sophisticated, they are increasingly involved in the inventive process:

• Drug discovery — AI systems like AlphaFold predict protein structures and generate novel drug candidates
• Materials science — Machine learning models design new alloys, polymers, and composite materials
• Engineering design — Generative design software creates optimized mechanical components and structures
• Algorithm development — AI systems create new algorithms, compression techniques, and optimization methods

If the law required that only humans can conceive inventions, but AI systems are doing the creative work, who owns the invention? Can it be patented at all? These questions have major implications for innovation policy and commercial value of AI-generated discoveries.

The DABUS Case: AI Inventorship Rejected Worldwide

The most prominent legal test of AI inventorship involved DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), an AI system created by Dr. Stephen Thaler. Thaler filed patent applications in multiple countries listing DABUS as the sole inventor, arguing that he could not truthfully claim to be the inventor himself because DABUS autonomously conceived the inventions.

Thaler v. Vidal (U.S. Federal Circuit, 2022)

The USPTO rejected Thaler's applications because DABUS was not a natural person. Thaler appealed to federal court, arguing that the Patent Act's use of "whoever" and "individual" should be interpreted to include AI systems.

The Federal Circuit's holding: The court affirmed the USPTO's rejection, ruling that under the Patent Act, an "inventor" must be a natural person. The court cited the statute's use of terms like "himself" and "herself" and the historical understanding that inventors are human beings. The Supreme Court declined to review the case in 2023, leaving the Federal Circuit's ruling as binding precedent.

International Rejections

Thaler filed similar applications worldwide, with nearly universal rejection:

• European Patent Office (EPO) — Rejected. "The inventor designated in a European patent must be a human being."
• United Kingdom — Initially rejected by UK IPO, affirmed by UK Supreme Court (2023): only natural persons can be inventors
• Australia — Initially allowed (2021), but reversed on appeal (2022): AI cannot be an inventor
• South Africa — Granted a patent listing DABUS as inventor (2021), but this decision is considered an outlier with little precedential weight

The near-universal conclusion: AI systems cannot be patent inventors under current law.

Current U.S. Law: Only Natural Persons Can Be Inventors

Statutory Language (35 U.S.C. § 100(f))

The Patent Act defines "inventor" as "the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention."

Courts have interpreted "individual" to mean a natural person — a human being.

Conception Requires a Human Mind

Patent law defines "conception" as the mental act of forming a definite and permanent idea of the complete and operative invention. Courts have long held that conception occurs in the mind of the inventor — and only humans have the requisite mental capacity and intent to "conceive" an invention in the legal sense.

Key principle: Inventorship is not about who physically builds or tests the invention — it's about who conceived the idea. Under current law, only humans can conceive inventions.

USPTO Guidance on AI-Assisted Inventions (2024)

In February 2024, the USPTO issued formal guidance addressing inventorship when AI systems assist in the inventive process. The guidance clarifies that while AI cannot be an inventor, humans who use AI as a tool can still be inventors if they make significant contributions to the conception of the invention.

Key Points from the 2024 USPTO Guidance

1. AI-assisted inventions are patentable — Using AI in the inventive process does not automatically disqualify an invention from patent protection.
2. At least one natural person must be named as inventor — Every patent application must list at least one human inventor who made a significant contribution to the invention.
3. Significant human contribution is required — The human inventor(s) must have contributed significantly to the conception of at least one claim. Merely recognizing a problem, providing data to an AI system, or owning the AI system is insufficient.
4. Truthful and accurate inventorship — Inventors must be listed accurately. Falsely claiming inventorship (or omitting an actual inventor) can invalidate a patent.

What Is NOT Sufficient for Inventorship

The USPTO guidance specifically states that the following alone do not make someone an inventor:

• Merely presenting a problem to an AI system
• Owning or having access to the AI system
• Providing data or training materials to the AI
• Recognizing the output of an AI system as valuable or useful
• Supervising the AI system's operation

These activities may be commercially valuable, but they do not constitute inventive contribution under patent law.

What Constitutes "Significant Human Contribution"?

The critical question for AI-assisted inventions is: What level of human contribution is sufficient to qualify someone as an inventor?

Examples of Significant Human Contribution

According to USPTO guidance, a person may be an inventor if they:

• Formulated the problem in a specific way that led the AI to a particular solution
• Designed or trained the AI system with the specific goal of solving a particular problem
• Selected and modified AI-generated output in a way that reflects significant creative contribution
• Recognized and appreciated the significance of AI output as solving a problem when the AI did not
• Contributed to the conception of a specific claim limitation that makes the invention patentable over prior art

The Pannu Factors (Adapted for AI)

Courts traditionally use the Pannu factors to determine inventorship in collaborations. These factors apply to AI-assisted inventions:

• Did the person contribute to the conception or reduction to practice of the invention?
• Did the person make a significant contribution to the claimed invention?
• Did the person contribute more than just well-known techniques or routine optimization?

If a person only provided routine input or direction to an AI system, and the AI autonomously generated the inventive concept, that person may not qualify as an inventor.

Gray Area: When AI Does Most of the Creative Work

The hardest cases involve AI systems that autonomously generate solutions with minimal human guidance. If a researcher simply asks an AI system to "design a new drug for diabetes" and the AI generates a novel compound, who is the inventor?

Under current law, the researcher must identify some significant contribution they made — such as:

• Defining specific constraints or parameters that guided the AI
• Selecting or modifying the AI's output based on inventive insight
• Training the AI with specific datasets or objectives

If the human's contribution was truly minimal and the AI operated autonomously, the invention may not be patentable for lack of a human inventor — even if the output is novel and useful.

Practical Implications for AI Companies and Researchers

Document Human Contributions

If you're using AI in the inventive process, document all human contributions contemporaneously:

• Record how you formulated the problem or research question
• Document the prompts, constraints, or parameters you provided to the AI
• Explain why you selected specific AI outputs over others
• Note any modifications or refinements you made to AI-generated results

This documentation will be critical for establishing inventorship and defending the patent against challenges.

Avoid Over-Claiming or Under-Claiming Inventorship

• Over-claiming: Listing someone as an inventor who did not contribute to conception (e.g., a manager who merely funded the project or a CEO who owns the AI system) can invalidate the patent.
• Under-claiming: Omitting a person who did contribute to conception (e.g., the data scientist who designed the AI model) can also invalidate the patent.

Work with a patent attorney to accurately identify inventors based on actual contributions to the claims.

AI as a Tool vs. AI as a Co-Inventor

Current law treats AI as a tool, similar to laboratory equipment or software. Just as a microscope or CAD program cannot be an inventor, neither can an AI system — even if it generates creative output.

However, as AI becomes more autonomous, this analogy may break down. If an AI system truly operates independently without human direction, the invention may fall into a legal gray area where no one can legitimately claim inventorship.

International Approaches to AI Inventorship

European Patent Office (EPO)

The EPO has rejected AI inventorship but allows AI-assisted inventions if at least one natural person contributed to the invention. EPO Guidelines state that "AI and machine learning tools can be used in the inventive process, but the invention must be made by a natural person."

United Kingdom

The UK Supreme Court ruled in 2023 that only natural persons can be inventors under UK patent law. However, the UK Intellectual Property Office has clarified that AI-assisted inventions are patentable if a human made a significant contribution.

Australia

Australia initially granted a patent listing DABUS as inventor in 2021, but this decision was reversed on appeal in 2022. The Federal Court held that an inventor must be a natural person. Australia now aligns with the U.S. and EPO approach.

China

China has not issued definitive guidance on AI inventorship, but Chinese patent law similarly requires inventors to be natural persons. Chinese courts and patent examiners appear to be following a similar approach to the U.S.: AI-assisted inventions are patentable if humans contributed significantly.

South Korea

South Korea amended its Patent Act in 2021 to explicitly state that AI-assisted inventions are patentable as long as a natural person contributed significantly to the conception. This clarification makes South Korea one of the more AI-friendly patent jurisdictions.

South Africa

South Africa remains the only jurisdiction to grant a patent listing AI as the inventor (DABUS patent granted in 2021). However, South Africa's patent system operates on a registration basis without substantive examination, so this decision carries limited precedential weight internationally.

Future Outlook: Proposed Legislation and Policy Debates

Proposed Patent Law Reforms

Several policy proposals have emerged to address AI inventorship:

• Employer/owner inventorship — Allow the owner or operator of the AI system to be listed as inventor when AI autonomously generates an invention
• Special AI patent category — Create a new type of patent protection specifically for AI-generated inventions with different requirements or shorter terms
• Public domain default — Allow AI-generated inventions with no significant human contribution to enter the public domain immediately
• Mandatory disclosure — Require patent applicants to disclose when AI was used in the inventive process

As of March 2026, no major jurisdiction has enacted legislation changing the natural person requirement for inventorship.

Policy Considerations

The debate over AI inventorship involves competing policy goals:

Arguments for allowing AI inventorship:

• Encourages AI development and use in research
• Ensures valuable inventions can be protected and commercialized
• Reflects the reality that AI systems are doing creative work
• Prevents "innovation gap" where AI-generated discoveries fall into public domain

Arguments against AI inventorship:

• Could flood patent system with low-quality AI-generated applications
• Reduces incentive for human innovation and research
• Raises questions about who can own or enforce AI-generated patents
• Patent law exists to incentivize human innovation, not AI operation

The USPTO and Congress continue to monitor AI developments and may issue further guidance or legislation as AI capabilities advance.

How to Handle Patent Applications for AI-Generated Inventions

Step 1: Identify All Human Contributors

Interview everyone involved in the project. Ask:

• Who formulated the problem or research question?
• Who designed, trained, or configured the AI system?
• Who selected or refined the AI's output?
• Who recognized the significance of the AI's results?

Only list as inventors those who made significant contributions to the conception of claimed subject matter.

Step 2: Document the Inventive Process

Create records showing:

• What problem was presented to the AI
• What constraints or parameters were set by humans
• How AI output was evaluated and selected
• What modifications were made by humans

This documentation supports the claim that humans contributed to conception.

Step 3: Draft Claims to Reflect Human Contributions

Work with a patent attorney to draft claims that highlight aspects of the invention where humans made creative contributions. Avoid claims that rest entirely on AI-autonomous output.

Step 4: Disclose AI Use in Prosecution

While not currently required, disclosing that AI was used in the inventive process demonstrates good faith and transparency. Some practitioners include this information in the specification.

Step 5: Monitor Evolving Law and Guidance

AI inventorship law is changing rapidly. Work with a patent attorney who stays current on USPTO guidance, court decisions, and international developments in this area.

Frequently Asked Questions

Can artificial intelligence be listed as an inventor on a patent?

No — under current U.S. law, only natural persons (human beings) can be named as inventors on a patent application. The Federal Circuit ruled in Thaler v. Vidal (2022) that AI systems like DABUS cannot be inventors because the Patent Act requires inventors to be individuals, interpreted as humans. This rule applies in nearly all major patent jurisdictions worldwide, including the EPO, UK, and Australia. South Africa is the only jurisdiction that has granted a patent listing AI as an inventor, but this is considered an outlier.

Can I patent an invention that was created by AI?

Yes, if at least one human made a significant contribution to the conception of the invention. The USPTO's 2024 guidance clarifies that AI-assisted inventions are patentable as long as a natural person contributed significantly to at least one claim. Simply presenting a problem to AI, owning the AI system, or recognizing AI output as valuable is not sufficient — the human must have contributed to the inventive concept through problem formulation, AI design/training, output selection/modification, or other creative input.

What is the DABUS case and why does it matter?

DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) is an AI system whose creator, Dr. Stephen Thaler, attempted to file patent applications worldwide listing DABUS as the sole inventor. The USPTO, Federal Circuit, EPO, UK courts, and most other jurisdictions rejected these applications, establishing that AI cannot be a patent inventor under current law. The case is significant because it tested the boundaries of inventorship law and confirmed that patents require human inventors — at least for now.

Who owns the patent if AI generates the invention?

Ownership depends on employment agreements and assignment contracts, just like any other invention. If a company employee uses AI to generate an invention, the company typically owns the patent rights through employment agreements (work-for-hire). If an independent inventor uses AI tools, they own the rights unless they've assigned them to someone else. The key is that at least one human must be named as inventor based on significant contribution — then ownership flows from that human inventor to employers, assignees, or other parties via contract.

Do I need to disclose that I used AI in my patent application?

Currently, the USPTO does not require disclosure of AI use in the inventive process. However, you must accurately identify all human inventors and cannot list AI as an inventor. Some patent attorneys recommend voluntarily disclosing AI involvement in the specification to demonstrate transparency and good faith. As the law evolves, disclosure requirements may be implemented. The most important legal obligation is to list inventors truthfully based on who actually contributed to conception of the claimed invention.

What happens if no human can legitimately claim inventorship?

This is the emerging legal gray area. If an AI system autonomously generates an invention with truly minimal human involvement — such that no person can honestly claim significant contribution to conception — the invention may not be patentable under current law. The invention would fall into the public domain because there is no valid inventor. This creates a potential "innovation gap" where valuable AI-generated discoveries cannot receive patent protection. Some policymakers argue this demonstrates the need for patent law reform; others argue it's an acceptable tradeoff to prevent AI-driven patent proliferation.

How are other countries handling AI inventorship?

Most major patent jurisdictions have reached the same conclusion as the U.S.: AI cannot be an inventor. The European Patent Office, UK, Australia, Canada, and Japan all require human inventors. South Korea amended its Patent Act to clarify that AI-assisted inventions are patentable if a natural person contributed significantly. China has not issued definitive guidance but appears to be following a similar approach. South Africa is the only jurisdiction that granted a patent listing AI as inventor, but this decision has little international influence. The trend is toward requiring human inventorship while allowing AI as a tool in the inventive process.

Warning & Disclaimer: The pages, articles, and comments on michaelmeyerlaw.com do not constitute legal advice, nor do they create any attorney-client relationship. The articles published express the personal opinions and views of the author as of the time of publication.