On December 31, 2024, China's National Intellectual Property Administration (CNIPA) released the Guidelines for Patent Applications for AI-Related Inventions (Trial Implementation), which clarifies the categorisation and patentability criteria for inventions involving artificial intelligence. This development follows a short comment period on a draft released just three weeks earlier, raising concerns about the adequacy of public consultation, particularly given that regulatory procedures typically stipulate a minimum comment period of 30 days.
The Guidelines identify four distinct types of AI-related patent applications. The first type covers inventions where artificial intelligence serves as an auxiliary tool in the invention process. An example outlined in the document illustrates this by noting instances where AI technology assists in the identification of protein binding sites, leading to the development of new drug compounds.
The second category pertains to field applications that employ AI algorithms across a variety of industries, including transportation, telecommunications, life sciences, security, commerce, education, entertainment, and finance. These applications aim to foster technological innovation and enhance operational efficiency.
The third category consists of functions based on AI algorithms, which include subfields such as natural language processing, computer vision, speech processing, knowledge representation and reasoning, and data mining. The document describes these functions as integral components that enable computational devices to perform tasks typically requiring human intelligence.
The final category involves patent applications that integrate AI algorithms or models into the invention itself. This includes solutions that leverage AI technology to achieve specific functionalities or to improve existing products and methods, such as a new type of electron microscope that utilises AI for image enhancement.
A significant aspect of the Guidelines is the assertion that AI systems cannot be recognised as inventors. As per the current legal framework, only natural persons who contribute creatively to the features of an invention can be credited as inventors on patent applications. The Guidelines explicitly state that terms like "artificial intelligence" cannot be included in the inventor section of patent documents, reinforcing that all named inventors must be human beings.
The document outlines criteria for subject matter eligibility, emphasising that claims need to demonstrate technical solutions that comply with natural laws to be eligible for patenting. Several scenarios illustrate acceptable technical solutions, including the use of AI algorithms to solve specific technical problems by processing data with direct technical significance or establishing a defined technical relationship between the AI model and computer system architecture.
The restrictions and nuances introduced by the Guidelines also extend to scenarios involving big data mining using AI algorithms. A claim must demonstrate how the inherent correlations identified conform to natural laws and produce valid technical effects; otherwise, it does not qualify as a technical solution.
An example provided pertains to a predictive method for food safety risks that constructs a knowledge graph based on dynamic data. This method incorporates timestamp information, allowing for more accurate risk prediction by reflecting the changing nature of food over time. In contrast, claims rooted solely in socio-economic correlations, rather than natural laws, fall outside the guidelines for being considered a technical solution.
The implementation of these Guidelines is poised to shape the landscape of AI-related inventions in China, providing clarity in a rapidly evolving field while continuing to adhere to existing legal principles regarding patentability and inventor recognition.
Source: Noah Wire Services