Artificial Intelligence and Inventorship: Global Perspectives on Machine Contributions to Innovation

Comparing US, European, and Asian Approaches to AI-Assisted Patent Applications

Category: Patent Law | Subcategory: International Patent Practice

Introduction: The Machine That Would Be Inventor

In 2018, a machine called DABUS—Device for the Autonomous Bootstrapping of Unified Sentience—generated two inventions without meaningful human intervention: a food container based on fractal geometry and a flashing light device designed to attract attention during emergencies. Dr. Stephen Thaler, DABUS's creator, then did something unprecedented: he filed patent applications in multiple jurisdictions naming DABUS itself as the sole inventor.

The applications triggered a global legal reckoning. Patent offices and courts from Washington to London to Beijing were forced to confront a question their governing statutes never anticipated: Can a machine be an inventor? More fundamentally, if artificial intelligence systems are increasingly capable of generating novel solutions to technical problems, what role must humans play in the inventive process for patent rights to attach?

The answers have varied dramatically across jurisdictions. The United States, European Patent Office, and United Kingdom have uniformly rejected AI inventorship, holding that inventors must be natural persons. But Australia briefly recognized AI inventorship before reversal on appeal, and South Africa granted a DABUS patent without substantive examination. More significantly, even among jurisdictions agreeing that inventors must be human, substantial disagreement exists about what quantum of human contribution suffices—a question with profound implications for companies deploying AI in research and development.

This divergence creates both uncertainty and opportunity. Companies using AI to accelerate innovation face genuine questions about whether their AI-assisted inventions are patentable, and if so, how to structure development processes and document human contributions to preserve patent rights across multiple jurisdictions. The stakes are enormous: artificial intelligence is increasingly central to pharmaceutical discovery, materials science, semiconductor design, and countless other fields where patent protection drives investment and competitive advantage.

This article provides a comprehensive analysis of global approaches to AI-assisted inventions, examining both the formal question of who can be named as inventor and the substantive question of what human contribution is required for valid inventorship. We compare approaches across the USPTO, EPO, UKIPO, and major Asian patent offices, analyze emerging guidance and case law, and provide practical strategies for companies navigating multinational patent protection in an environment of legal uncertainty. For related discussion of AI patentability considerations, see our analysis of AI-generated inventions and ownership.

The Fundamental Question: What Is an Inventor?

Traditional Inventorship Doctrine

Patent law has historically assumed that inventors are human beings exercising creative faculties to solve technical problems. This assumption was so foundational that statutes rarely defined "inventor" at all—the term's meaning seemed self-evident.

In the United States, the Patent Act defines an inventor as "the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention." 35 U.S.C. § 100(f). The statute uses "individual" rather than "person," suggesting a natural person rather than a legal entity. Courts have interpreted inventorship to require conception—"the formation in the mind of the inventor of a definite and permanent idea of the complete and operative invention." Townsend v. Smith, 36 F.2d 292, 295 (C.C.P.A. 1930). This mental act, by definition, requires a mind.

The European Patent Convention similarly contemplates human inventors, providing that the inventor has the right to be mentioned in the European patent. Article 62 EPC. The implementing regulations require that the application designate the inventor, and the EPO has interpreted this to require a natural person with legal capacity.

These traditional formulations assumed a world where machines were tools—sophisticated instruments that extended human capabilities but did not themselves conceive solutions. A chemist might use computer modeling to predict molecular behavior, but the chemist conceived the invention. An engineer might use CAD software to refine a design, but the engineer directed the creative process. The machine was always means, never mind.

The AI Disruption

Artificial intelligence, particularly machine learning and neural networks, disrupts these assumptions by enabling machines to identify solutions that humans did not conceive, direct, or sometimes even understand. A neural network trained on chemical compound data might identify a novel molecule with therapeutic properties through pattern recognition that no human chemist directed or anticipated. A generative design system might produce mechanical structures optimized for specific constraints through evolutionary algorithms, yielding solutions no human engineer envisioned.

In these scenarios, the traditional conception-based test for inventorship becomes problematic. If conception requires formation in a human mind, and no human mind formed the inventive idea, who is the inventor? The AI system itself? The humans who created the AI? The humans who trained it on particular data? The humans who defined the problem the AI addressed? Or is there simply no inventor—and therefore no valid patent?

The DABUS applications forced patent offices worldwide to confront these questions explicitly. The answers reveal fundamental differences in how jurisdictions conceptualize inventorship and the patent system's purposes.

United States: Natural Persons Only, But Questions Remain

The DABUS Litigation

The USPTO rejected Dr. Thaler's DABUS application in April 2020, holding that the Patent Act requires inventors to be natural persons. The rejection relied on the statutory text—"individual" in § 100(f)—and the Federal Circuit's prior holding that inventors must be natural persons, not corporations. Beech Aircraft Corp. v. EDO Corp., 990 F.2d 1237, 1248 (Fed. Cir. 1993). The USPTO's decision stated: "Because the Patent Act's text, consistent with our precedent, supports the conclusion that an inventor must be a natural person and the petition does not dispute that the inventor is not a natural person, the petition is denied."

Dr. Thaler challenged this rejection in federal court. In Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), the Federal Circuit affirmed the USPTO's rejection. The court applied the Chevron framework, finding that the Patent Act unambiguously requires human inventors. The statutory term "individual" ordinarily means a natural person, the court reasoned, and nothing in the Patent Act's context suggested a different meaning. The court rejected arguments that policy considerations—promoting innovation, adapting to technological change—should override textual analysis, holding that such policy arguments are properly directed to Congress, not courts.

The Supreme Court declined to hear the case, leaving the Federal Circuit's holding as controlling law: under current U.S. patent law, inventors must be natural persons. AI systems cannot be named as inventors on U.S. patent applications.

The Unresolved Human Contribution Question

The DABUS litigation resolved the formal inventorship question—AI cannot be named as inventor—but left open the more practically significant question: What human contribution suffices for valid inventorship when AI plays a substantial role in the inventive process?

The USPTO has provided some guidance. In February 2024, the USPTO published guidance on AI-assisted inventions addressing this question. The guidance confirms that while AI cannot be an inventor, "the use of an AI system by a natural person(s) does not preclude a natural person(s) from qualifying as an inventor." The key inquiry is whether a natural person made a "significant contribution" to the claimed invention.

The guidance adopts the existing Pannu factors for joint inventorship, asking whether the natural person:

1. Contributed in some significant manner to the conception or reduction to practice of the invention
2. Made a contribution to the claimed invention that is not insignificant in quality, when measured against the full scope of the claimed invention
3. Did more than merely explain well-known concepts or the current state of the art

Pannu v. Iolab Corp., 155 F.3d 1344, 1351 (Fed. Cir. 1998).

Applying these factors to AI-assisted inventions, the USPTO guidance indicates that a natural person who merely presents a problem to an AI system, without more, likely does not qualify as an inventor. But a natural person who designs, builds, or trains the AI system in a manner specifically directed to elicit a particular solution may qualify. Similarly, a natural person who recognizes the significance of an AI output, refines it, or contributes additional inventive elements may be a proper inventor.

Practical Implications for U.S. Filings

The USPTO's guidance creates both requirements and opportunities for companies using AI in innovation:

Documentation Requirements: Companies must document human contributions throughout the development process to support inventorship claims. This documentation should identify: who defined the problem the AI addressed; who selected or designed the AI system used; who curated training data; who interpreted AI outputs; who recognized the inventive significance of AI suggestions; and who contributed additional elements beyond what the AI generated.

Process Structure: Development processes should be structured to ensure meaningful human involvement at multiple stages. Purely autonomous AI invention—where humans merely pose a problem and accept AI output—likely cannot support valid inventorship. But collaborative human-AI processes, where humans direct, interpret, and refine AI contributions, can support human inventorship.

Claim Drafting: Patent claims should be drafted to emphasize elements reflecting human contribution. If humans contributed specific technical insights while AI contributed others, claims emphasizing the human contributions may be more defensible than claims emphasizing AI contributions.

The following table summarizes the U.S. approach:

| Issue | U.S. Position | |-------|---------------| | Can AI be named as inventor? | No—inventors must be natural persons | | Legal basis | 35 U.S.C. § 100(f); Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022) | | Human contribution required? | Yes—"significant contribution" to conception or reduction to practice | | Guidance source | USPTO AI Inventorship Guidance (Feb. 2024) | | Key factors | Pannu factors for joint inventorship |

European Patent Office: Consistent Rejection, Evolving Guidance

The EPO's DABUS Decisions

The European Patent Office rejected DABUS applications through both first-instance and appeals proceedings, establishing clear precedent that inventors under the European Patent Convention must be natural persons.

The EPO's Receiving Section initially rejected the DABUS applications under Rule 19(1) EPC for failure to properly designate an inventor. On appeal, the Legal Board of Appeal affirmed in decisions J 8/20 and J 9/20 (December 2021). The Board held that the EPC's reference to "inventor" presupposes a natural person, relying on: the ordinary meaning of "inventor" as a person who invents; the EPC's provisions regarding the inventor's right to be mentioned (Article 62) and the transfer of rights from inventor to applicant (Article 60), which contemplate legal relationships only natural persons can hold; and the systematic context of patent law, which has historically assumed human inventors.

The Board explicitly rejected arguments that recognizing AI inventors would better serve the patent system's purposes, holding that such policy considerations could not override the EPC's textual requirements. If AI inventorship is to be recognized, the Board indicated, it must come through amendment to the EPC—not judicial interpretation.

The EPO's Approach to AI-Assisted Inventions

While rejecting AI inventorship, the EPO has addressed AI-assisted inventions through its examination guidelines and evolving practice. The EPO's approach emphasizes that the use of AI as a tool does not preclude patentability, provided human inventors make contributions sufficient to support inventorship.

The EPO Guidelines for Examination state that computer-implemented inventions, including those using AI and machine learning, are patentable if they meet the standard patentability requirements: novelty, inventive step, industrial applicability, and sufficient disclosure. The guidelines address AI-specific issues including:

Technical Character: AI and machine learning methods must have technical character to be patentable. A mathematical method or algorithm per se lacks technical character. But AI applied to solve a technical problem—controlling a technical system, processing technical data, analyzing physical phenomena—may acquire technical character through its application.

Inventive Step: For AI inventions, the inventive step analysis considers whether the claimed technical effect would have been obvious to the skilled person. The "skilled person" is assumed to have access to common general knowledge, including knowledge of standard AI/ML techniques. Therefore, merely applying known AI techniques to a known problem is likely obvious; the inventive contribution must lie elsewhere—in the training data selection, the AI architecture, the application to a new technical field, or another technical insight.

Enablement: The specification must enable a skilled person to reproduce the AI system. For machine learning inventions, this may require disclosure of training data characteristics, network architecture, training methodology, and other information necessary to achieve the claimed results.

Human Contribution Under EPO Practice

The EPO has not issued formal guidance equivalent to the USPTO's February 2024 publication specifically addressing the quantum of human contribution required for AI-assisted inventions. However, the EPO's existing inventorship doctrine provides a framework.

Under EPC Article 60, the right to a European patent belongs to the inventor or their successor in title. The inventor is the natural person who made the inventive contribution—the creative act that distinguishes the invention from the prior art. For AI-assisted inventions, this inquiry asks: Did a natural person make the creative contribution, using AI as a tool? Or did the AI itself make the creative contribution, with humans merely implementing or recognizing it?

EPO practice suggests that human contributions at various stages can support inventorship: conceiving the technical problem; selecting or designing the AI system to address the problem; selecting and preparing training data; interpreting AI outputs and recognizing their significance; and refining or improving AI-generated solutions. The key is whether the human's contribution rises to the level of an inventive act—a creative contribution that would not be obvious to the skilled person.

United Kingdom: Legislative Clarity After DABUS

The UK DABUS Litigation

The United Kingdom became a key battleground for AI inventorship through the DABUS litigation, ultimately producing clear appellate guidance rejecting AI inventors.

The UK Intellectual Property Office (UKIPO) initially rejected Dr. Thaler's DABUS applications on grounds that the designated inventor was not a "person" as required by the Patents Act 1977. The High Court upheld this rejection, finding that the statutory scheme contemplates human inventors. On appeal, the Court of Appeal affirmed in Thaler v. Comptroller-General of Patents [2021] EWCA Civ 1374, holding that an inventor must be a natural person under UK law.

The Supreme Court granted permission to appeal, and in December 2023 issued its decision in Thaler v. Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49. The Supreme Court unanimously rejected Dr. Thaler's appeal, holding that:

1. Under Section 7 of the Patents Act 1977, an "inventor" must be a natural person. The term has always been understood to mean a human being, and nothing in the statutory context suggests Parliament intended otherwise.

2. Dr. Thaler could not derive rights from DABUS to apply for a patent because DABUS, as a machine, cannot hold property rights capable of transfer.

3. Policy arguments about promoting AI innovation, while potentially meritorious, are matters for Parliament to address through legislation, not for courts to address through statutory interpretation.

The Supreme Court's decision provides authoritative guidance that UK patent law, as currently written, requires human inventors. Only legislative amendment could change this requirement.

UKIPO Guidance on AI-Assisted Inventions

Following the DABUS litigation, the UKIPO has provided guidance on AI-assisted inventions that parallels but does not precisely mirror the USPTO's approach. The UKIPO acknowledges that AI systems are increasingly used as tools in the inventive process and that such use does not preclude patentability, provided natural persons make inventive contributions.

The UKIPO's approach emphasizes identifying the "actual deviser" of the invention—the person whose intellectual effort produced the inventive concept. For AI-assisted inventions, this inquiry asks whether a natural person contributed the inventive insight, even if AI systems contributed to implementation or optimization.

Key considerations under UKIPO practice include:

Conception vs. Implementation: The inventive contribution lies in conception—the idea—not merely implementation. A person who conceives an inventive solution and uses AI to implement or optimize it may be the inventor. But a person who merely presents a problem to an AI system and accepts its output may not have contributed the inventive conception.

Problem Definition: Defining a technical problem, while valuable, may not alone constitute inventive contribution. The inventorship inquiry focuses on the solution, not merely identifying what needs to be solved.

Recognition and Selection: Recognizing the significance of an AI-generated output—identifying it as novel, useful, and worth pursuing—may contribute to inventorship if such recognition involves technical judgment rather than mere observation.

Asian Perspectives: Divergent Approaches Across Major Economies

China: Pragmatic Flexibility

China has not directly confronted DABUS-style applications but has developed guidance on AI-assisted inventions that reflects pragmatic attention to the country's substantial AI industry.

The China National Intellectual Property Administration (CNIPA) has indicated that inventors must be natural persons under Chinese patent law, consistent with the Patent Law's requirement that inventor designations identify the "person" who made the inventive contribution. Article 17 of China's Patent Law Implementing Regulations requires naming the inventor, understood as a natural person.

However, China's approach to AI-assisted inventions appears more flexible regarding the quantum of human contribution required. CNIPA guidance suggests that human involvement in defining problems, selecting AI tools, directing AI processes, and evaluating outputs can support inventorship even when AI systems generate substantial portions of technical solutions.

Chinese practice has also grappled with disclosure requirements for AI inventions. The CNIPA has emphasized that AI/ML inventions must be sufficiently disclosed to enable reproduction, including information about algorithms, training data, and model architectures. Specifications that merely describe AI outputs without enabling their generation may be rejected for insufficient disclosure.

Japan: Emphasis on Human Creativity

The Japan Patent Office (JPO) has taken a position consistent with other major offices: inventors must be natural persons. The JPO rejected the premise that AI systems can be designated as inventors, interpreting Japan's Patent Act to require human inventors.

Japanese patent law emphasizes that inventive activity requires human creative acts. Article 2(1) of Japan's Patent Act defines "invention" as the "highly advanced creation of technical ideas utilizing the laws of nature." The JPO interprets this to require human intellectual contribution—machines cannot engage in "creation" within the statute's meaning.

For AI-assisted inventions, the JPO's approach focuses on whether humans contributed the creative act that constitutes the invention. Using AI as a tool—even a very sophisticated tool—does not preclude human inventorship if humans directed the creative process. But if AI systems autonomously generated the inventive concept without meaningful human intellectual contribution, no valid inventor exists.

The JPO has provided guidance through its examination guidelines addressing AI/ML inventions, focusing primarily on patentability requirements (novelty, inventive step, enablement) rather than inventorship specifically. These guidelines recognize AI/ML as legitimate technical fields while applying standard patentability analysis.

South Korea: Following International Consensus

The Korean Intellectual Property Office (KIPO) has aligned with the international consensus that inventors must be natural persons. KIPO has not faced DABUS applications directly but has indicated that Korean patent law contemplates human inventors.

Korean practice for AI-assisted inventions emphasizes documentation of human contributions. KIPO guidance suggests that applicants should clearly identify: the natural persons who contributed to the invention; the nature of each person's contribution; how AI systems were used as tools in the inventive process; and the relationship between human conception and AI-generated outputs.

KIPO's examination guidelines address AI/ML inventions within the broader framework of computer-implemented inventions, applying standard novelty, inventive step, and enablement analysis while recognizing AI-specific considerations for disclosure sufficiency.

South Africa: The Outlier

South Africa became a notable outlier when it granted a DABUS patent in July 2021—the first jurisdiction to issue a patent naming an AI system as inventor. The grant occurred because South Africa's patent system uses a formal examination (depository) system rather than substantive examination. Applications that meet formal requirements are granted without examination of patentability or proper inventorship.

The South Africa DABUS grant does not reflect a policy determination that AI systems can be inventors under South African law. Rather, it reflects the absence of substantive examination that would have addressed the inventorship question. The grant has been criticized as anomalous and potentially invalid if challenged.

Australia: Reversal After Initial Recognition

Australia produced the most dramatic jurisprudential development in the AI inventorship debate when a federal court initially recognized AI inventorship before reversal on appeal.

In Thaler v. Commissioner of Patents [2021] FCA 879, Justice Beach of the Federal Court of Australia held that DABUS could be named as inventor under Australia's Patents Act 1990. Justice Beach reasoned that the Act does not define "inventor" and does not explicitly require human inventors. He found that interpreting "inventor" to include AI systems better served the patent system's purposes of promoting innovation.

This decision made Australia briefly unique among major jurisdictions in recognizing AI inventorship. However, the Full Federal Court reversed in Commissioner of Patents v. Thaler [2022] FCAFC 62. The Full Court held that "inventor" in the Patents Act necessarily means a natural person, based on the word's ordinary meaning, the statutory context, and the historical understanding of inventorship. The High Court of Australia denied special leave to appeal, leaving the Full Federal Court's human-inventor requirement as settled law.

The Australian reversal illustrates the international consensus: across common law and civil law jurisdictions, appellate courts have consistently held that patent law requires human inventors. Legislative change would be required to recognize AI inventorship.

The Substantive Question: What Human Contribution Suffices?

A Spectrum of Human Involvement

While jurisdictions agree that inventors must be natural persons, they have provided less clarity on the degree of human contribution required when AI systems participate substantially in the inventive process. Human involvement in AI-assisted invention exists on a spectrum:

Full Human Conception: At one end, humans conceive the complete invention and use AI only for implementation—running simulations, optimizing parameters, generating production-ready designs. Here, human inventorship is unproblematic; AI serves as a sophisticated tool no different conceptually from a calculator or CAD system.

Human Direction with AI Generation: Humans define the problem, select the AI system, specify constraints, and interpret outputs, but AI generates the technical solution. Humans exercise substantial judgment throughout but do not themselves conceive the inventive solution.

Minimal Human Involvement: Humans pose a problem to an AI system and accept its output with minimal intermediate involvement. The AI autonomously identifies the inventive solution; humans merely recognize its value.

Autonomous AI Invention: At the far end, AI systems identify problems and generate solutions without meaningful human direction. This is the DABUS scenario—and the one jurisdictions have clearly rejected as supporting no valid inventorship.

The practical challenge lies in the middle of this spectrum, where human involvement is substantial but AI contributes significant creative elements. Most real-world AI-assisted invention falls here, and jurisdictions are only beginning to provide guidance.

Factors Affecting Human Contribution Analysis

Emerging guidance from the USPTO, EPO, and other offices suggests several factors relevant to assessing whether human contributions suffice for inventorship:

Problem Definition and Selection: Identifying a technical problem to solve is valuable but may not alone constitute inventive contribution. The inventorship inquiry focuses on the solution's conception, not merely problem identification.

AI System Design and Selection: Choosing or designing an AI system specifically suited to a problem may contribute to inventorship if the selection reflects technical judgment about how to achieve the inventive result. But selecting an off-the-shelf AI system without modification likely does not.

Training Data Curation: For machine learning systems, selecting and preparing training data can involve substantial technical judgment that contributes to inventorship. Curating training data to enable the AI to generate particular types of solutions may support inventorship claims.

Constraint Specification: Specifying the constraints within which an AI system operates—performance requirements, physical limitations, optimization objectives—may contribute to inventorship if the constraints embody technical insights about the desired solution.

Output Interpretation and Recognition: Recognizing the significance of an AI-generated output—identifying it as novel, useful, and inventive—may contribute to inventorship if such recognition requires technical expertise rather than mere observation.

Refinement and Improvement: Taking an AI-generated output and refining it through additional human insight clearly supports inventorship, at least with respect to the refined elements.

Integration into Larger Systems: Incorporating an AI-generated component into a larger human-designed system may support inventorship of the larger system, even if the component itself reflects primarily AI contribution.

The following table summarizes factors affecting human contribution analysis:

| Factor | Likely Supports Inventorship | Likely Insufficient Alone | |--------|------------------------------|---------------------------| | Conceiving complete solution, using AI for implementation | Yes | — | | Designing AI system specifically for problem | Possibly | — | | Curating training data with technical judgment | Possibly | — | | Specifying technically meaningful constraints | Possibly | — | | Recognizing significance of AI output | Possibly | Possibly | | Refining AI output through additional insight | Yes | — | | Integrating AI output into larger system | Yes (for larger system) | — | | Merely posing problem to AI | — | Probably | | Accepting AI output without modification | — | Probably |

Documenting Human Contributions: Practical Strategies

Building an Inventorship Documentation Program

Companies using AI in research and development should implement systematic documentation practices to preserve inventorship options for AI-assisted inventions. Key elements include:

Contemporaneous Records: Document human contributions as they occur, not after inventions are identified. Contemporaneous records—lab notebooks, development logs, meeting minutes—provide stronger evidence than retrospective summaries.

Contributor Identification: For each development project, identify all individuals who contributed technical insight. Track not only who performed work but who contributed inventive ideas.

Contribution Characterization: For each contributor, document the nature of their contribution: Did they define the problem? Select the AI system? Curate training data? Interpret outputs? Refine solutions? Identify commercial significance?

AI Role Documentation: Document what the AI system contributed and, equally important, what it did not contribute. If humans made key technical decisions that shaped the AI's output, document those decisions.

Decision Points: Identify decision points where humans exercised technical judgment—choosing among AI-generated alternatives, modifying AI suggestions, combining AI outputs with human insights.

Structuring Development Processes

Beyond documentation, companies can structure development processes to ensure meaningful human contribution:

Human-in-the-Loop Design: Design AI-assisted development processes with human decision points at critical stages. Rather than fully autonomous AI generation, implement processes where humans direct, evaluate, and refine AI outputs.

Technical Review Requirements: Require technical review of AI-generated outputs before adoption. Reviewers should apply technical judgment, not merely accept or reject.

Iterative Refinement: Structure development as iterative collaboration between humans and AI systems, where humans provide input, AI generates outputs, humans evaluate and refine, and the cycle continues.

Domain Expert Involvement: Ensure domain experts—not just AI specialists—participate in AI-assisted development. Domain expert contributions may be more clearly inventive than generic AI operation.

Inventor Interviews and Declarations

When preparing patent applications for AI-assisted inventions, conduct thorough inventor interviews to identify and document human contributions:

Contribution Mapping: For each claim element, identify which human(s) contributed the underlying concept and how AI systems were involved.

Conception Documentation: Document when and how each human contributor conceived their contributions. Conception evidence supports inventorship claims.

Declaration Preparation: Prepare inventor declarations that accurately characterize each inventor's contribution, including their relationship to AI systems used in development.

Consistency Review: Review documentation for consistency across inventors and with technical records. Inconsistencies may undermine inventorship claims.

Multinational Filing Strategies

Jurisdiction-Specific Considerations

Companies filing AI-assisted inventions internationally must consider jurisdiction-specific requirements while maintaining consistent global positions:

United States: Apply the USPTO's significant contribution standard and Pannu factors. Document human contributions sufficient to demonstrate conception or significant contribution to reduction to practice. Consider whether claims emphasize elements reflecting human contribution.

Europe: Ensure specifications meet EPO enablement requirements, including sufficient disclosure of AI system architectures, training methodologies, and other information necessary for reproduction. Prepare to address technical character requirements at examination.

United Kingdom: Following the Supreme Court's DABUS decision, apply the "actual deviser" standard. Document the natural person whose intellectual effort produced the inventive concept.

China: Meet CNIPA disclosure requirements for AI/ML inventions. Chinese examination may focus more on enablement than inventorship, but maintain documentation supporting human inventorship.

Japan: Emphasize human creative contribution consistent with JPO interpretation of "invention" as human creation. Document the human intellectual acts that produced the claimed subject matter.

Korea: Align with KIPO documentation expectations identifying human contributors and their roles.

Coordinating Global Applications

Multinational filing strategies should address:

Consistent Inventorship: Identify inventors consistently across jurisdictions. While legal standards may vary, factual determinations about who contributed what should be consistent.

Specification Harmonization: Prepare specifications that meet the most demanding disclosure requirements (likely EPO or China for AI inventions) and use them as the basis for all filings.

Priority Claims: Establish priority through timely first filings, typically in jurisdictions with favorable treatment of AI-assisted inventions.

Claim Adaptation: While maintaining consistent claim scope where possible, consider jurisdiction-specific claim adaptations that emphasize elements meeting local patentability standards.

Prosecution Coordination: Coordinate prosecution across jurisdictions to develop consistent positions and avoid inconsistent admissions.

Subject Matter Eligibility Considerations for AI Inventions

The Section 101 Challenge in the United States

Beyond inventorship, AI-assisted inventions face patentability challenges under subject matter eligibility doctrine. In the United States, Section 101 of the Patent Act excludes "abstract ideas" from patentability. AI and machine learning methods, involving mathematical algorithms and data processing, frequently face Section 101 rejections.

The USPTO applies a two-step framework from Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014): First, determine whether claims are directed to a patent-ineligible concept (abstract idea, law of nature, natural phenomenon). Second, if so, determine whether claim elements, individually or as an ordered combination, transform the claim into patent-eligible subject matter by adding an "inventive concept."

AI claims directed to algorithms per se, mathematical calculations, or data processing without technical application typically fail Step 1. Even claims applying AI to specific problems may fail if the application is viewed as merely implementing an abstract idea on generic computer components.

Strategies for overcoming Section 101 rejections include:

Technical Integration: Claim AI systems as integrated into technical processes—controlling physical devices, processing sensor data, producing concrete technical outputs.

Specific Technical Improvements: Emphasize specific technical improvements over prior art, such as faster processing, reduced computational resources, or improved accuracy for specific technical applications.

Non-Generic Implementation: Avoid claims that could be characterized as implementing abstract ideas on generic hardware. Claim specific AI architectures, training methodologies, or technical implementations.

Practical Application: Demonstrate practical application of AI methods to technical problems, moving beyond mere data processing to technical solutions.

EPO Technical Character Requirements

The EPO applies a different framework but reaches similar results. Mathematical methods, including AI algorithms, lack "technical character" and are excluded from patentability under Article 52(2) EPC. However, AI methods applied to technical problems may acquire technical character through their application.

The EPO examines whether claimed AI contributes to a "technical effect" beyond mere data processing. AI controlling technical systems, processing data representing physical measurements, or producing technically applicable outputs may achieve technical character. AI merely processing abstract data without technical application likely does not.

For inventive step analysis, the EPO's "COMVIK approach" considers only technical features when assessing obviousness. Non-technical features (including abstract algorithmic concepts) may be in the claim but do not contribute to inventive step. AI inventions must demonstrate inventive step through technical contributions, not merely novel algorithms.

Looking Forward: Legislative and Policy Developments

Calls for Legislative Reform

The DABUS litigation has prompted calls for legislative reform to address AI inventorship. Advocates argue that denying patent protection for AI-generated inventions may:

• Discourage disclosure of AI innovations (companies may choose trade secret protection instead)
• Create uncertainty about ownership of AI-generated innovations
• Disadvantage jurisdictions that reject AI inventorship relative to those that might accept it
• Fail to incentivize the development of beneficial AI systems

Counterarguments emphasize that:

• The patent system exists to incentivize human innovation through exclusive rights
• AI systems do not need patent incentives to "invent"
• Recognizing AI inventorship would create difficult questions about ownership, duration, and enforcement
• Existing frameworks can accommodate AI-assisted invention through human inventorship

Policy Studies and Consultations

Multiple patent offices and legislative bodies have undertaken studies of AI and intellectual property:

USPTO: The USPTO conducted public consultations on AI and inventorship, publishing reports analyzing stakeholder views. The USPTO's October 2020 "Public Views on Artificial Intelligence and Intellectual Property Policy" report found divided views on whether AI inventorship should be recognized but majority support for maintaining human inventorship requirements.

WIPO: The World Intellectual Property Organization has convened conversations on AI and IP policy, exploring whether international frameworks should address AI inventorship. WIPO's work has focused on information-gathering and dialogue rather than proposing specific reforms.

European Commission: The European Commission has studied AI and intellectual property, including whether existing frameworks adequately address AI-generated innovations. Commission reports have not recommended recognizing AI inventorship but have called for continued monitoring.

UK Government: Following the DABUS litigation, the UK government consulted on AI and intellectual property, ultimately concluding that no immediate legislative change was warranted but committing to ongoing review.

Potential Future Frameworks

Several alternative frameworks have been proposed for addressing AI-generated innovations:

Modified Inventorship: Recognize AI contributions while requiring human involvement—perhaps allowing AI to be named as co-inventor alongside human inventors, or creating a new category of "AI-assisted inventor."

Assignee-Based Registration: Allow registration of AI-generated innovations based on the entity controlling the AI system, without requiring inventor designation. This would parallel existing frameworks for computer-generated works in some copyright systems.

Reduced Protection Terms: Provide protection for AI-generated innovations but with shorter terms than standard patents, reflecting the reduced human investment in their creation.

Disclosure Without Exclusivity: Require disclosure of AI-generated innovations (to advance the public knowledge) without granting exclusive rights, or with more limited exclusivity than traditional patents.

Sui Generis Protection: Create a new form of intellectual property protection specifically for AI-generated innovations, with requirements and terms tailored to AI's characteristics.

None of these alternatives has gained significant traction, and the immediate future likely involves continued application of existing human-inventorship frameworks with evolving guidance on AI-assisted invention.

Practical Checklist for AI-Assisted Patent Applications

Pre-Filing Assessment

Before filing patent applications for AI-assisted inventions, companies should assess:

• [ ] Has at least one natural person made a significant contribution to the claimed invention?
• [ ] Can the nature and extent of human contributions be documented?
• [ ] Does documentation support human conception or significant contribution to conception?
• [ ] Are human contributions sufficient under the standards of intended filing jurisdictions?
• [ ] Have all contributing humans been identified for potential inventorship?
• [ ] Are there elements of the invention to which AI contributed autonomously without meaningful human input?
• [ ] Can claims be drafted to emphasize elements reflecting human contribution?

Documentation Checklist

Maintain documentation addressing:

• [ ] Problem definition: Who identified the technical problem? When? Based on what insight?
• [ ] AI system selection: Who chose the AI system used? Why that system?
• [ ] AI system design/modification: Did humans design or modify the AI system for this problem?
• [ ] Training data: Who selected/curated training data? What technical judgments were involved?
• [ ] Constraint specification: Who specified constraints for AI operation? What technical insights informed constraints?
• [ ] Output evaluation: Who evaluated AI outputs? What technical criteria were applied?
• [ ] Output selection: If multiple outputs, who selected among them? On what technical basis?
• [ ] Refinement: Did humans refine or improve AI outputs? How?
• [ ] Integration: How were AI-generated elements integrated into the overall invention?

Application Preparation

When preparing applications:

• [ ] Conduct thorough inventor interviews mapping contributions to claim elements
• [ ] Prepare specifications meeting most demanding disclosure requirements (for global filing)
• [ ] Include sufficient AI system disclosure for enablement (architecture, training, data characteristics)
• [ ] Draft claims emphasizing technically inventive elements with human contribution
• [ ] Prepare inventor declarations accurately characterizing contributions
• [ ] Consider jurisdiction-specific claim adaptations
• [ ] Maintain documentation supporting inventorship determinations

Conclusion: Navigating Uncertainty Through Documentation and Process

The global patent system has reached a clear consensus on the formal question of AI inventorship: inventors must be natural persons. No major jurisdiction recognizes AI systems as inventors, and recent appellate decisions in the United States, Europe, United Kingdom, and Australia have foreclosed judicial recognition of AI inventorship. Only legislative change could alter this requirement.

But the formal consensus masks substantial uncertainty on the practical question that matters most: what human contribution suffices when AI systems play substantial roles in generating inventions? The USPTO has provided guidance applying the Pannu factors; the EPO and UKIPO have offered less specific direction; and Asian offices have largely deferred to general inventorship principles. As AI becomes more capable and more central to research and development, the pressure for clearer standards will intensify.

Companies deploying AI in innovation cannot wait for legal clarity. They must act now to preserve patent options for AI-assisted inventions by documenting human contributions, structuring development processes to ensure meaningful human involvement, and preparing applications that accurately characterize the human-AI collaboration producing claimed inventions.

The stakes extend beyond individual patent applications. How patent systems ultimately accommodate AI-assisted innovation will shape incentives for AI development, the balance between disclosure and secrecy in AI research, and the competitive positions of companies and countries in AI-intensive industries. For now, the path forward runs through careful documentation of human contributions within existing frameworks—a practical strategy that preserves options while legal and policy debates continue.

Related Articles

• AI-Generated Inventions: Who Owns What the Machine Creates?
• Copyright Infringement Claims Against Generative AI: The New York Times, Getty, and What Comes Next
• Trade Secrets in the Age of Remote Work and Cloud Computing
• Open Source Licensing Landmines in Enterprise Software Development
• Data Scraping After HiQ v. LinkedIn: Copyright, Contract, and Computer Fraud Claims
• The Right of Publicity Meets Digital Doubles: Deepfakes, AI Avatars, and Celebrity Likeness

This article is intended for informational purposes only and does not constitute legal advice. Readers should consult with qualified patent counsel regarding their specific circumstances.

Practice Contacts:
[Partner Name], Patent Prosecution & Strategy — [email]
[Partner Name], International Patent Practice — [email]
[Associate Name], AI & Emerging Technologies — [email]