Specialist Chapter: How to Assess Patent Eligibility of AI Inventions in China

In summary

In China, patent eligibility of AI inventions is assessed by examining in sequence whether the AI inventions fall under the rules and methods for mental activities and whether they constitute a technical solution. An AI invention claim can pass the examination as to mental activities as long as it contains at least a technical feature. The ‘three elements of technology’ test is then applied to examine whether the claim constitutes a technical solution, requiring the claim to contain some technical means applying the laws of nature to solve a certain technical problem, with some technical effects achieved.

Discussion points

• AI invention
• Patent eligibility
• Algorithmic features
• Technical features
• Mental activities
• Technical solutions

Referenced in this article

• Article 2, paragraph 2 of the file:///Users/isabel.holme/Downloads/
• Patent Law of the People’s Republic of China
• Article 25, paragraph 1, item 2 of the Patent Law of the People’s Republic of China
• Part II, Chapter 9, section 6 of the revised ‘Guidelines for Patent Examination’ (2010)
• China National Intellectual Property Administration

Artificial intelligence (AI) technology is currently undergoing revolutionary development. Generally, ‘artificial intelligence’ can be defined as ‘theories, methods, technologies and application systems that utilize digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge, and use the knowledge to obtain optimal results’^[1]. Since AI technology is centred around the simulation of how humans perceive and process information, like hearing and vision, it has been innovatively applied to fields such as voiceprint recognition, facial recognition, driverless cars, intelligent customer service chatbots, machine translation and medical image processing.

The essence of AI technology is algorithms. As a result, most, if not all AI inventions are solutions involving algorithms, although algorithms per se are explicitly excluded, as rules and methods for mental activities, from patent-eligible subject matters under Chinese patent law.^[2] With patent applications relating to new AI technologies surging, the question of how to assess the patent eligibility of AI inventions has become an important issue of concern to the Chinese patent industry as well as to innovation subjects.

Rules for examining patent eligibility of AI inventions

In the revised ‘Guidelines for Patent Examination’, which were issued by the China National Intellectual Property Administration (CNIPA) on 31 December 2019 and entered into force on 1 February 2020, a new section titled ‘Provisions on the examination of patent applications for invention containing algorithmic features or features of business rules and methods’ was added for patent applications for inventions relating to AI, Internet Plus, big data and blockchain.^[3] This new section provides detailed rules for examining patent applications for inventions relating to these topics and aims to standardise the examination criteria of such applications.

According to this new section of the revised ‘Guidelines for Patent Examination’, the patentability examination of patent applications for AI inventions will be conducted in order as follows:

• examination of patent eligibility; and
• examination of novelty and inventiveness.

Specifically, for the examination of patent eligibility, whether a claim of a patent application falls under the rules and methods for mental activities as stipulated in article 25, paragraph 1, item 2 of the Chinese Patent Law (article 25.1 (2)) will first be examined. If the claim, considered as a whole, does not fall under the rules and methods for mental activities, then the examination will proceed to determine whether it constitutes a technical solution as referred to in article 2, paragraph 2 of the Chinese Patent Law (article 2.2).

The CNIPA emphasises that the eligibility examination of patent applications for AI inventions shall follow such criteria that the examination shall be carried out on the solution for which the patent protection is sought, meaning the solution defined by the claim. The examination of such solution shall be conducted in a way that ensures all of the contents recorded in the claim are taken as a whole to analyse the technical means involved, the technical problems solved and the technical effects obtained, instead of simply breaking the claim down into technical features and algorithmic features or features of business rules and method (BM features), which are then evaluated separately.

In particular, in the examination under article 25.1(2), if a claim contains one or more technical features in addition to algorithmic features or BM features, the claim, viewed as a whole, is deemed not to fall under the rules and methods for mental activities as stipulated in article 25.1(2), and should not be excluded from patent-eligible subject matters. The claim is considered to fall under the rules or methods for mental activities only if it is drawn to just an abstract algorithm, or simply BM features, and does not contain any technical features. For example, a mathematical modelling method based on an abstract algorithm that does not contain any technical features falls into the rules and methods for mental activities, and is thus ineligible for patent protection.

In the examination under article 2.2, it is necessary to consider all the features recited in the claim as a whole and apply the ‘three elements of technology’ test. According to this test, if the claim contains some technical means that apply the laws of nature to solve a technical problem and thereby achieves some technical effects in compliance with the laws of nature, then the claimed solution constitutes a technical solution as referred to in article 2.2. In practice, a claim on an AI invention can pass the examination under article 2.2, provided that the AI algorithm recited in the claim is applied in a specific technical field to solve a technical problem with some technical effects obtained. Illustratively, the AI invention will be deemed as a technical solution if the steps concerning the algorithm recited in the claim are each closely related to the technical problem to be solved (eg, the data processed by the algorithm is data with a concrete technical meaning in the technical field), if the execution of the algorithm can directly reflect the process of applying the laws of nature to solve a certain technical problem and if some technical effects can be achieved.

Examples of patent eligibility examination of AI inventions

Example 1

A claim drawn to a model training method reads as follows.

A method for training a model comprising a first sub-model, a second sub-model and a third sub-model, the method comprising:

• obtaining training samples comprising a labelled sample set and an unlabelled sample set; and
• training the first sub-model, the second sub-model and the third sub-model using the training samples to obtain the trained model.

For the examination of the patent eligibility of the claim, firstly it will be examined to check whether the claimed subject matter falls under the rules and methods for mental activities as stipulated in article 25.1(2), and if it does not, then it will be examined to check if it constitutes a technical solution as referred to in article 2.2.

In this example, the claim contains only algorithmic features, so it would be rejected as falling under the rules and methods for mental activities as stipulated in article 25.1(2).

Example 2

The claim of Example 1 is redrafted to read as follows.

A method applied to a computer for training a model comprising a first sub-model, a second sub-model and a third sub-model, the method comprising:

• obtaining training samples from a labelled sample set and an unlabelled sample set stored in a storage space; and
• training the first sub-model, the second sub-model and the third sub-model using the training samples to obtain the trained model.

In this example, the claim contains some technical features like ‘a computer’ and ‘a storage space’ in addition to the algorithmic features, so it shall not be rejected as falling under rules and methods for mental activities as stipulated in article 25.1(2).

However, the claim of Example 2 fails to be applied in a specific technical field to solve a technical problem. In particular, the objects processed by the algorithmic steps – ‘a labelled sample set’, ‘an unlabelled sample set’, ‘training samples’ and ‘model’ – are all abstract mathematical concepts rather than data with a concrete technical meaning in the technical field, and the execution of the algorithm fails to directly reflect the process of applying the laws of nature to solve a certain technical problem with any technical effects obtained. Therefore, it would fail to meet the ‘three elements of technology’ test, so the claimed method in Example 2 is not a technical solution as referred to in article 2.2.

Example 3

The claim of Example 1 is then redrafted to read as follows.

A method applied to a computer for training a model adapted for detecting internet abnormal access behaviour and comprising a first sub-model, a second sub-model and a third sub-model, the method comprising:

• obtaining training samples from a sample set labelled as abnormal access data or as normal access data and an unlabelled sample set that are stored in a storage space; and
• training the first sub-model, the second sub-model and the third sub-model using the training samples to obtain the model for detecting internet abnormal access behaviour.

The claimed method of Example 3 is directed to solve a technical problem of how to detect internet abnormal access behaviour in the technical field of the internet. The objects processed by the algorithmic steps, such as ‘a sample set labelled as abnormal access data or as normal access data’ and ‘the model for detecting Internet abnormal access behaviour’, are data with a concrete technical meaning in the relevant technical field, and the execution of the algorithm directly reflects the process of applying mathematics that belongs to the laws of nature to solve the technical problem of how to detect internet abnormal access behaviour with certain technical effects obtained. Therefore, the claimed method of Example 3 passes the ‘three elements of technology’ test as it constitutes a patent-eligible technical solution, as referred to in article 2.2.

Conclusion

From the above examples, it can be understood that according to the current practice of examining the patent eligibility of AI inventions in China, a relatively low-threshold criterion is used in the examination of whether a claim falls under the rules and methods for mental activities as stipulated in article 25.1(2). As long as the claim contains a technical feature or technical features, it can usually pass the examination. However, the “three elements of technology” test, which is a relatively high-threshold criterion, is also applied to examine whether a claim constitutes a technical solution as referred to in article 2.2. This test requires the claim to contain some technical means applying the laws of nature to solve a certain technical problem, with some technical effects in compliance with the laws of nature achieved thereby.

Footnotes