Artificial Intelligence (AI), defined as systems capable of performing tasks traditionally requiring human intelligence, is transforming industries such as autonomous vehicles, healthcare diagnostics, smart manufacturing and telecommunications. The rapid advancement and commercial importance of AI have intensified the need to protect AI-driven inventions through patents.
However, patenting AI inventions presents various challenges since in many jurisdictions, patent office classify algorithms, mathematical methods, and abstract ideas as excluded subject matter, making it critical for AI-related inventions to demonstrate a particular technological application.
Another significant challenge with patenting AI-related inventions is strict and varied requirement of patent offices globally. The patent offices across various jurisdictions require that the patent application must sufficiently and clearly describe a system architecture, data processing methods, training procedures, and how the AI achieves a technical effect. Generic references to AI models without detailed technical information are unlikely to meet the standards of patentability.
Further, differences in statutory laws and practice across various jurisdictions play a significant role in shaping the approach to patenting the AI-driven inventions. Several countries have issued recent updates to examination guidelines, refining the criteria for subject matter eligibility and disclosure sufficiency for AI-based inventions. Understanding these jurisdiction-specific nuances is essential to securing robust and enforceable protection for the AI-driven inventions.
United States
In the U.S., the patent eligibility of AI inventions is governed by 35 U.S.C. § 101, which defines eligible subject matter as “any new and useful process, machine, manufacture, or composition of matter.” However, abstract ideas, including mathematical algorithms, remain ineligible unless they demonstrate a practical application.
The United States Patent and Trademark Office (USPTO) issued 2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence, effective July 17, 2024, which clarifies the application of the Alice/Mayo framework to AI inventions. This guidance emphasises that claims must demonstrate a practical application and technological improvement beyond abstract computation. Further, the guidance includes Examples 47–49 which illustrate how AI-related claims are analyzed for eligibility. The guidance is incorporated into the USPTO’s subject matter eligibility policy and is integrated in the Manual of Patent Examining Procedure (MPEP).
The USPTO’s Inventorship Guidance for AI-Assisted Inventions (effective February 13, 2024) and Federal Register notice further clarify that only natural persons can be named as inventors, even if AI tools contributed to the invention. It is essential that the human inventor’s role is substantial and well-documented in the patent application, with AI tools treated as assistive rather than inventive entities.
Europe
The European Patent Office (EPO) evaluates AI-related inventions based on Article 52 of the European Patent Convention (EPC), which excludes “programs for computers” and “mathematical methods” from patentability unless they produce a technical effect. The EPO Guidelines for Examination specify that AI inventions must demonstrate measurable improvements in performance of a system, device, or hardware beyond mere data processing. or information presentation.
The EPO carefully examines claims that are centred around abstract computational tasks, such as organizing data, performing classifications, or using predictive algorithms that do not tangibly affect a system. Claims must be system-oriented, showing how the AI functionality is integrated into a technical process or device. The patent application must disclose the technical contribution and system-level benefits, including details of the AI model, training dataset, AI model structure, system architecture, and reproducible training and validation methods. The EPO examiners look for detailed explanations showing how the AI technology contributes to solving a practical technical problem in a real-world system.
Australia
Under Section 18 of the Patents Act 1990, a threshold for patentable subject matter is that an invention must be a manner of manufacture. For AI-related inventions, this means more than simply running an algorithm on a generic computer i.e., the claimed invention must provide a technical contribution beyond abstract computation and produce a physical effect or practical transformation. For an AI invention to be patentable in Australia, the patent application must demonstrate evidence of measurable improvements in operation of a system, device, or process, such as faster processing, more reliable system performance, or better use of system resources.
In Australia, claims that are limited to algorithms, data analysis, or abstract modeling without integration into a practical system typically do not satisfy the criteria for patentable subject matter. The detailed disclosure must include AI architecture, system-level deployment, training methodologies, evidence of the physical effect or operational improvement. This approach aligns with global best practice and is essential for demonstrating manner of manufacture in the AI context.
Further, Australia’s Patent Manual of Practice and Procedure makes it clear that an inventor must be a natural person. Listing an AI system or a non-human entity as an inventor will contravene the requirements of the Act and will be considered an issue of non-compliance with formality requirements.
New Zealand
New Zealand applies a strict approach to computer-implemented inventions under the Patents Act 2013. For AI inventions to be patent eligible, they must demonstrate a technical contribution that goes beyond normal computer functions and must produce a tangible effect that enhances system or device performance. Abstract AI concepts, standalone data processing, or predictive algorithms that do not directly impact a system are typically excluded from patent protection in New Zealand.
To be eligible, AI-related inventions must be applied within a system or method that solves a specific technical problem existing in conventional arts. Moreover, the patent application must present comprehensive details, including system structure, how the AI model is integrated, the flow of data, and the technical advantages achieved as a result. The examiners in New Zealand prioritize AI inventions that materially enhance hardware or system functionality, rejecting claims that are limited to computational processes without real-world application.
China
Under Article 25 of the Chinese Patent Law, “rules and methods for intellectual activities” (including abstract algorithms) are not eligible for patent protection. However, the China National Intellectual Property Administration (CNIPA) Guidelines allow AI inventions that are embedded in technical solutions and deliver tangible technical improvements. For this, AI must be implemented within a functional system or module.
Further, the patent application must also show clear, quantifiable improvements, such as enhanced processing speed, higher efficiency, or increased automation. Additionally, the patent application should include a thorough explanation of data flows, system layouts, and the detailed design of individual components and modules.
India
Section 3(k) of the Indian Patent Act, 1970 excludes “a mathematical or business method or a computer programme per se or algorithms” from patentability. However, AI inventions are eligible if they demonstrate a technical application and are integrated into a system or hardware. The invention must follow a problem-solution approach and must be linked to a particular system or operational method. The application must disclose the system architecture, hardware integration, and the operational role of the AI. The invention must address a specific technical challenge and must be firmly linked to a particular system or operational method.
As of now, India has not issued any specific guidelines for examining AI inventions and instead relies on Guidelines for Examination of Computer Related Inventions (CRIs), 2017 to evaluate the patentability. Further, the subject matter exclusions stated in Section 3(k) of the Act are applied to assess patentability of the AI inventions.
Conclusion
Patenting AI inventions requires strict adherence to the legal and procedural requirements of each jurisdiction. For an AI invention to qualify for patent protection globally, it must be applied within a practical system or process, and the technical advancement must be thoroughly and specifically described. The application must provide detailed, reproducible disclosure of AI architecture, training data, and system integration and broad or vague claims must be avoided with focus on specific technical improvements and real-world applications. Further, it is ensured that inventorship is properly attributed to natural persons, with human contributions clearly documented.
The USPTO 2024 Guidance has introduced a more structured and demanding approach to evaluating AI inventions, while jurisdictions such as Europe, China, India, Australia, and New Zealand each follow their own legal standards and examination criteria. By understanding and complying with these laws and guidelines, inventors can secure enforceable, valuable protection for their AI innovations worldwide.
