Prosecuting AI-related patents in China

Allowable claim formalities

AI technologies are closely linked to software applications, often manifesting as computer programs. In China, both method and product claims are permissible for computer program innovations. Product claims may include systems, computer apparatus and computer-readable storage mediums. There can also be AI-related hardware innovations, which are also potentially protectable. Allowable claim formalities include:

• Method Claims:

• A method for [function], comprising:

• Step A;
• Step B;
• Step C;
• Step D.

• System Claims:

• A system for [function], comprising:

• Means (Module) for A;
• Means (Module) for B;
• Means (Module) for C;
• Means (Module) for D.

• Computer Apparatus Claims:

• A computer apparatus comprising (a) processor(s) and (a) memory(s), with a computer program stored on the memory(s), wherein the computer program, when executed by the processor(s), implements the following steps:

• Step A;
• Step B;
• Step C;
• Step D.

• Computer-Readable Storage Medium Claims:

• A computer-readable storage medium with a computer program stored on it, wherein the computer program, when executed by the processor(s), implements the following steps:

• Step A;
• Step B;
• Step C;
• Step D.

• Computer Program Product Claims:

• A computer program product including computer program(s) (instructions), wherein the computer program(s)/instructions(s), when executed by the processor(s), implement the following steps:

• Step A;
• Step B;
• Step C;
• Step D.

Patentable subject matter

Article 2.2 of the Patent Law and Article 25.1 of the Patent Law exclude abstract results of human intellectual activity from the subject of patentability. Article 2.2 of the Patent Law provides that an invention refers to a new technical solution proposed for a product, a process or an improvement thereof. Article 25.1 of the Patent Law provides that no patent shall be granted for: ... (2) Rules and methods for human intelligence activities. 

In practice, examiners usually will first review whether the claimed invention falls into the scope of rules and methods for human intelligence activities of Article 25.1 (the Article 25 review). If the claimed invention does not fall into that scope, the examiner then will review whether it is a technical solution prescribed in Article 2.2 (the Article 2 review).

The Article 25.1 Review

According to examination practices, if all the technical features defined by a claim are solely abstract processing rules or simple mathematical methods, and are not combined with any specific actual application, the claimed solution is regarded as a rule and method of human intelligence activity and is thus subject to non-patentability under Article 25.1 of the Patent Law. Even if the field of application is mentioned, if the claim does not reflect the application and relevance of the algorithm in the specific field, it cannot pass the Article 25 review. Some specific points are listed below for reference.

The Article 2 Review

After the Article 25 review, examiners will further consider whether the technical solution belongs to the technical solution stipulated in Article 2.2 of the Patent Law. That is whether the overall technical solution uses technical means that conform to the laws of nature to solve technical problems and achieve certain technical effects. 

From a patent examination perspective, the use of natural laws is generally reflected in the following three aspects: 1) using physical laws of industrial processes, measurement or testing process control; 2) dealing with external technical data that conform to the laws of nature, such as image data, flow rate data, and even quantity distribution data; 3) using the laws of nature to improve the internal performance of the computer, such as reducing the number of computer hardware resources called by adjusting parameters. Some specific points below for reference:

Sufficiency and enablement

The "black box" nature of AI technology can conflict with the full disclosure requirement of patent law. The deduction process between input and output in AI is often opaque, making it challenging to describe accurately. However, there can be innovations in the specific architecture of AI models, training approaches, innovative control of hardware / software, and it is important to describe these in as much detail as possible to both explain how the components interact or are interconnected, and how the improvements can be achieved.

Article 26.3 Review

Article 26.3 of the Patent Law stipulates that the specification shall clearly and completely describe the invention or utility model, which shall be subject to the realization of the invention by a person skilled in the technical art. If the specification of a patent application only states the application of artificial intelligence, the inputs, and the corresponding outputs, it is often considered that the disclosure of the specification is insufficient.

In view of the objective existence of the black box feature, the Chinese National Intellectual Property Administration has listed the elements that should be possessed for different types of AI inventions in its Guidelines, which we summarize as follows:

Article 26.4 Review

Article 26.4 of the Patent Law provides that the claims shall be based on the description and clearly and concisely define the scope of the claimed patent protection. Applicants may deploy several sets of claims to comprehensively protect rights, and applicants must imbed sufficient contents supporting these claims into the description. Some points for reference: