In summary
This article examines how patent systems in the United States, Europe and China are addressing the legal and practical challenges posed by AI-related innovations. It analyses key doctrines including inventorship attribution, subject matter eligibility and disclosure requirements, with a focus on how each jurisdiction interprets the role of human contribution and technical character in artificial intelligence-related applications. While all three regions exclude artificial intelligence (AI) systems from being named inventors, they permit protection for AI-assisted inventions if human involvement is clearly established and technical criteria are met. Through comparative analysis and practical recommendations, this article aims to support cross-border patent prosecution for emerging technologies involving computational creativity.
Discussion points
•Inventors must be natural persons under US, EU and Chinese law; AI cannot be named.
•Eligibility standards diverge: section 101 in the US, ‘technical character’ in Europe and ‘technical solution’ in China.
•The EPO’s 2025 Guidelines require a technical contribution via implementation or application.
•China emphasises quantifiable technical effects and hardware integration in disclosures.
•Disclosure standards vary: the US focuses on requirement under section 112, Europe on technical clarity, and China on reproducibility and integration.
•Effective claim drafting must align with local thresholds and examination practices.
Referenced in this article
•Thaler v Vidal 43 F.4th 1207 (Fed. Cir. 2022)
•EPO Decision J 8/20
•US Code Title 35, sections 101 and 112
•China Patent Law Article 2(2) and Article 16
•CNIPA Guidelines for Examination, Part I, Chapter 1, section 4.1.2
•EPO Guidelines for Examination (April 2025)
Artificial intelligence (AI) is increasingly capable of generating novel and valuable outputs in fields such as chemistry, engineering and software development. This technological advancement has prompted fundamental questions in patent law: can AI be qualified as the inventor when it contributes to the inventive process? Are such AI-assisted or AI-generated innovations patentable? If so, what strategies should applicants adopt to ensure compliance with existing legal requirements during drafting and prosecution?
This article examines how three key jurisdictions – the United States, Europe, and China – are addressing these issues through evolving legal doctrines and administrative practices. It analyses how each system is responding to the challenges of inventorship, subject matter eligibility, and disclosure obligations in the age of AI-driven innovation.
Inventorship in the ‘Age of AI’: who can be credited?
The concept of inventorship has long been tied to the act of a human. With AI systems increasingly performing functions once reserved for human inventors, patent offices have had to clarify the boundaries between machine assistance and genuine inventorship. This section examines each jurisdiction’s stance on whether AI can be credited for inventorship.
United States: human contribution remains paramount
US law requires that inventorship be attributed to natural persons. This principle was recently reaffirmed by the Federal Circuit in Thaler v Vidal, 43 F.4th 1207 (Fed. Cir. 2022), where the court held that an AI system cannot be listed as an inventor under current US patent statutes. The ruling emphasised that an ‘inventor’ must be an ‘individual’, meaning a human being. As a result, while AI-assisted inventions are patentable, there must be a demonstrable human contribution to the conception of the invention. It is advisable for practitioners to clearly identify the natural persons who made a substantive contribution to the claimed invention.
Europe: inventorship and the role of the human mind
The European Patent Office (EPO) has likewise confirmed that inventorship must be limited to natural persons. In decision J 8/20, the EPO’s Legal Board of Appeal rejected applications that named the AI system DABUS as the inventor, affirming that inventorship under European Patent Convention (EPC) practice must refer to a human being (article 60(1) and decision J 8/20). This position aligns with the European legal doctrine that emphasises the role of human mental activity in the inventive process. Therefore, applicants should ensure that the invention is attributed to an identifiable human contributor, particularly in filings that involve substantial algorithmic development.
China: the line between tool and creator
China’s approach is grounded in article 16 of the Patent Law, which affirms that inventors have the right to be named in the patent application. Rule 14 of the Implementing Regulations clarifies that the inventor must be a natural person. Accordingly, the Guidelines for Patent Examination explicitly state that names of organisations, collectives or AI systems shall not be listed as inventors. While AI can assist in various stages of innovation, it is treated as a tool, not a creator.
Subject matter eligibility: what types of AI innovations qualify?
Even where a human inventor is present, AI-related inventions must satisfy each jurisdiction’s threshold for subject matter eligibility. This often depends on whether the claimed invention produces a technical effect, avoids abstraction, or qualifies as a technical solution.
United States: the section 101 barrier and the quest for an inventive concept
The United States applies the Alice/Mayo framework to determine subject matter eligibility under 35 USC section 101. Courts evaluate whether the claims are directed to an abstract idea, and if so, whether they embody a ‘technological improvement’ or an inventive concept that transforms the abstract idea into a patent-eligible application. AI inventions reciting conventional machine learning techniques without technical adaptation are likely to fail this test. To increase the likelihood of eligibility, applicants should emphasise improvements to computer functionality or the technical field of application. For instance, demonstrating how the AI model enhances processing speed, resource efficiency or system reliability may help overcome section 101 rejections.
Europe: defining ‘technical character’ through implementation and application
The EPO’s Guidelines for Examination (April 2025) emphasise that AI-related inventions must involve a technical contribution to be patentable. This technical character can arise through two dimensions:
•specific technical implementation: when the AI model is specifically adapted to improve the functioning of a computer or system component, such as optimising execution time or memory usage; and
•technical application: when the AI model solves a problem in a specific technical domain, such as medical imaging or signal processing.
Generic statements such as ‘controlling a technical system’ do not establish technical character. Applicants should present a clear link between the AI model and a concrete technical effect and frame the invention using the EPO’s problem-solution approach.
China: the ‘technical solution’ requirement
Eligibility under article 2(2) of the Patent Law requires that the claimed invention constitute a technical solution, which may be demonstrated through:
•processing technically defined data, such as sensor outputs in industrial control systems;
•interaction between the algorithm and hardware or system architecture that yields measurable performance improvements (eg, 20 per cent faster inference via GPU optimisation); or
•mining data relationships to solve technical problems in fields governed by natural laws (eg, biomedical pattern recognition).
Applications based solely on economic or social data patterns, without transformation grounded in technical principles or interaction with physical systems, do not meet the eligibility threshold.
Enabling disclosure: how much detail is enough?
Given the opacity of many AI models, patent offices from various jurisdictions increasingly emphasise the need to disclose how inputs lead to outputs in a reproducible and technically verifiable manner. To obtain valid protection, applicants must provide sufficient disclosure to enable skilled persons to reproduce the invention without undue experimentation. While grounded in traditional legal principles, these disclosure requirements are applied differently across jurisdictions.
United States: section 112 disclosure for AI systems
While the US Patent and Trademark Office (USPTO) has not issued AI-specific disclosure rules, AI-related inventions must comply with the enablement and written description requirements under 35 USC section 112. This includes:
•the model architecture (eg, neural network structure, activation functions);
•the training methodology (eg, data sources, preprocessing steps, hyperparameter ranges);
•integration with technical systems; and
•deployment configuration and output interpretation.
Mere functional claiming or reference to generic AI models is typically insufficient. To satisfy section 112, the application must describe the invention in a way that allows a person skilled in the art to make and use it without undue experimentation.
Europe: sufficiency and the technical effect nexus
The EPO mandates that AI-related inventions be disclosed with enough detail to allow reproducibility. Essential elements include the type of training data, data processing methods, model architecture, training steps and technical effect achieved. If the technical effect depends on characteristics of the data or model parameters, those must be disclosed. Applicants should avoid describing the AI system as a ‘black box’ and ensure that at least one embodiment fully supports the claim scope.
China: structured expectations for technical clarity
Chinese guidelines emphasise clear disclosure of an algorithm’s structure, training process, hardware integration, and technical outcomes. Vague references without specifying algorithmic structure or technical integration are insufficient. Working examples or flow diagrams are recommended to demonstrate reproducibility under article 26(3). The applicant must also identify how the human inventor devised the algorithmic features or supervised their integration into a technical framework.
Comparative analysis: converging principles, diverging paths
Despite differences in doctrinal emphasis and statutory frameworks, China, Europe and the United States converge on a foundational principle: only natural persons may be named as inventors. Each jurisdiction rejects the notion of AI as an autonomous inventor, aligning with longstanding legal traditions that tie inventorship to human intellect and responsibility.
Where the jurisdictions diverge, however, is in their approach to subject matter eligibility and disclosure. Europe continues to emphasise the need for a ‘technical contribution’, as assessed through the problem-solution framework. In the United States, subject matter eligibility under section 101 remains a major hurdle, particularly where AI claims are viewed as abstract without sufficient ‘inventive concept’. While both Europe and China require that an invention constitutes a ‘technical solution’, their respective thresholds differ. China’s approach emphasises quantifiable technical effects achieved via hardware integration or domain-specific application of natural laws – such as system resource allocation or data governed by physical principles. In contrast, the EPO focuses on whether the invention improves computer functionality or addresses a problem within an applied technical field.
Nevertheless, a global trend is emerging that supports the patentability of AI-assisted inventions, provided that a human is substantively involved in the inventive process and the invention integrates into a technical system. This convergence suggests that while procedural thresholds may differ, the core principles of inventive contribution, technical specificity, and human accountability remain intact. For patent professionals, success increasingly depends on jurisdiction-specific tailoring, backed by a strategic understanding of each region’s evolving stance on AI.
Practical drafting strategies: adapting to global patent systems
AI-related inventions require thoughtful claim drafting to navigate varying requirements across borders. While human inventorship and technical effect are universal requirements, the structure and focus of patent claims must be tailored to local legal doctrines.
Best practices for global applications
•Name a human inventor with identifiable contributions.
•Emphasise how the invention solves a technical problem.
•Include technical implementation details (eg, hardware, training method, system integration).
•Avoid abstract or purely algorithmic claims.
Strategic drafting approaches to avoid subject matter eligibility challenges in the US, Europe and China
In all jurisdictions, claims that describe only abstract concepts or general-purpose algorithms are at risk. Patent professionals should aim to:
•emphasise improvements to computer functionality and avoid generic AI claims in the United States;
•use the problem-solution framework to highlight technical effects in Europe; and
•quantify the technical effect of the AI model (eg, ‘such as by reducing signal classification error by 12%’), and provide specific details on how the algorithm integrates with hardware or system architecture in China.
Conclusion
AI-assisted inventions continue to test the boundaries of established patent doctrines. While current frameworks in the United States, Europe and China uniformly require that inventorship be attributed to natural persons, all three jurisdictions have developed pathways for protecting innovations that incorporate algorithmic tools – provided that human input remains central and the claimed invention satisfies jurisdiction-specific technical and disclosure standards.
For practitioners, success in prosecuting such applications increasingly depends on tailored drafting strategies that align with local examination practices. This includes careful articulation of technical contribution, quantifiable effects, and the inventor’s role in framing or interpreting algorithmic outputs. As the legal treatment of AI-generated innovation continues to evolve, patent professionals must remain vigilant and adaptive – grounding their strategy not only in emerging examination trends, but also in a principled understanding of what constitutes true inventive contribution in the age of computational creativity.
.jpg)