In response to the needs of innovators to further clarify the examination rules for patent applications in new industry conditions and new fields concerning AI, the China National Intellectual Property Administration issued Announcement 343 on 31 December 2019 regarding amendments to the Patent Examination Guidelines, which make explicit provisions for the examination of invention patent applications concerning AI, Internet+, big data and blockchains, among other things.
In terms of the drafting of AI-related invention patent applications and the response to any subsequent office actions, the following three issues are of major concern:
- sufficient disclosure; and
- novelty and inventiveness.
Relevant eligibility provisions
The innovation points of most AI-related inventions lie in the improvement of algorithms or software, which decide that the AI-related claims are composed mostly of technical features and non-technical features (algorithm features). In the examination process, if an AI-related claim includes only non-technical features (algorithm features) while lacking any technical features, or it includes both technical and non-technical features but any correlation between them is absent, the claim may face the challenge of eligibility.
The provisions in China’s Patent Law relating to the eligibility of AI patents include:
- whether the features of one claim are the rules and methods for mental activities (Article 25.1); and
- whether those features constitute a technical solution (Article 2.2).
In accordance with the amended Patent Examination Guidelines, the examiner will first determine whether the claim satisfies Article 25.1, and further determine whether Article 2.2 is met after determining that Article 25.1 is satisfied.
According to Article 25.1, if a claim includes only features relating to algorithms and rules for mental activities, instead of any technical features, it will probably be excluded from the subject matter of patent protection (exclusion clause). On the other hand, if a claim includes both algorithm features and technical features, and the claim as a whole is not a rule or method for mental activities, the possibility of patent right grant is, according to Article 25.1, not excluded.
The key point of Article 2.2 lies in the technicality of the solution. The amended guidelines stipulate that: “all the features recorded in one claim shall be considered as a whole. If the claim records that technical means utilising natural laws are adopted for the technical problem to be solved, and thus a technical effect conforming to the natural laws is obtained, the solution defined in the claim belongs to the technical solution stipulated in Article 2, paragraph 2 of the Chinese Patent Law.”
How to avoid the eligibility issue
Avoid algorithm-only claims: It is a basic premise of Article 25.1 that an AI-related claim should include both technical and non-technical features (algorithms). In general, mere algorithms are likely to be excluded from the subject matter of protection in the examination phase.
Consider application scenarios for non-technical features: Since the improvement point of an AI-related invention usually lies in the improvement of algorithms or software, while an algorithm or software can be applied to different scenarios to solve different problems, the description of the application scenarios is important to emphasise the technicality of an AI-related invention patent application, either in the phase of drafting the description and claims, or in the phase of responding to the office action. The definition of the application scenarios will play a key role in solving the issue of eligibility.
Pay attention to the correlation between technical and non-technical features: In the drafting phase and the phase of responding to the office action, attention should be paid to the correlation between technical features (eg, parameters in the application scenarios) and the non-technical features (algorithms). The correlation is the premise for the examiner to consider the two types of feature as a whole. In other words, even if an AI-related claim includes both the technical and non-technical features, the whole technical solution does not reflect the correlation between the two types of feature, and it remains a risk that the claim is not considered eligible (see Table 1).
Table 1: case study on eligibility
|Examination opinions and response strategies|
|Office action||Related claims||Defect||Response|
|Office action 1||Claims 1 to 7 (method)||Eligibility issue (Article 2.2)||Amendment|
|Claims 8 to 15 (program)||Eligibility issue (Article 25)||Amendment|
|Office action 2||Claims 1 to 16||Eligibility issue (Article 2.2)||Granted after argument|
A method for generating a two-class classification or prediction model (see Tables 2 and 3) comprises:
- obtaining a discriminant function for classifying a training sample;
- calculating a discriminant score for each training sample;
- determining whether the training sample is correctly classified;
- determining a misclassified sample region based on X;
- constructing a new training sample set by extracting said training samples;
- repeating points 1 to 5 for the new training sample set; and
- storing as a two-class classification or prediction model for samples.
Table 2: office action 1
|1 to 7 (methods)||Eligibility issue (Article 2.2)||A technical element is added to the subject.|
A method for generating a two-class classification or prediction model → a method for generating a compound toxicity prediction model.
|8 to 15 (programs)||Eligibility issue (Article 25)||The computer program is amended into a device corresponding to the method claim.|
A program for generating a two-class classification or prediction model → a device for generating a compound toxicity prediction model.
Table 3: office action 2
|1 to 16||Eligibility issue (Article 2.2)||Argument|
Examiner’s opinion: The invention adopts a mathematical method that essentially solves the classification prediction of a mathematical model, which has no technical meaning and does not constitute a technical problem.
Application scenarios for non-technical features: The subject of the invention is a method (device) for generating a compound toxicity prediction model. The ‘compound’ is a natural substance and the ‘compound toxicity’ is an inherent property of each compound, which exists objectively and cannot be obtained by a mathematical calculation based on the artificially specified principle.
Correlation between technical and non-technical features: The toxicity of a compound, such as whether it has mutagenicity, is determined by the chemical structure of the compound itself, and is an inherent characteristic determined by the structure of the compound. Therefore, the invention is based on a compound toxicity prediction model (emphasising the correlation between the compound toxicity and the prediction model), which utilises and is limited by natural laws.
Analysis and conclusion
As can be seen from the above case, the original claim is for an algorithm. The application scenario and technical features of the algorithm are not recorded in the claim during the drafting process, which is the main reason why the examiner doubts the eligibility issue. In the response process, a specific application scenario (technical attribute feature: compound toxic) is added, and the correlation between the natural attribute of the technical feature and the algorithm feature (ie, non-technical feature) is argued, finally overcoming the eligibility issue.
Sufficient disclosure (Article 26. 3)
The improvement point of an AI-related invention usually lies in an algorithm or software, which can be applied to different scenarios. An AI-related invention, however, will describe how it is embodied in a specific application scenario in addition to the overall application scenarios at the initial drafting phase. If the description offers only less specific algorithms and generic application scenarios, and persons skilled in the art cannot embody or obtain a deterministic result based on the technical solution recorded in the description, it will be deemed a ‘black-box operation’. In this situation, the problem of ‘insufficient disclosure’ will be encountered in the examination process.
To avoid insufficient disclosure, a number of points should be focused on during the drafting phase.
Specific application scenarios
If the final purpose of executing the computer program is to solve the problem of specific scenario applications, such as improving the user experience (eg, saving transaction costs or time) or improving management platform efficiency (eg, increasing consumption demand or creating a new service mode), on the one hand, it is necessary to clearly and explicitly state what better specific scenario application effect the invention can achieve in the summary of the description. On the other hand, it should also describe in detail the specific implementations capable of obtaining advantageous effects under the specific scenario applications in the embodiments.
The specific scenario application must be based on a certain technical problem. The technical features adopted to solve the technical problem will be interrelated and work together with the algorithm features adopted to solve the problem of the specific scenario application necessary to complete the whole process of the solution, instead of solving the corresponding problems individually and being pieced together into one solution. In other words, the technical field, the problem to be solved, the solution, the technical effect and the embodiment in the description must clearly define the problem and effect of the specific scenario application.
Essential technical parameters
The core of the invention patent application related to the AI technology is the solution recorded in the description. The hardware and software environments of the computer recorded in the description must be sufficient to implement the algorithm features in the solution. The description should clearly and completely describe the solution adopted by the invention to solve its technical problem or the problem of the specific scenario application. The solution should specify the technical means adopted to solve the above problems, or the mixed means containing both the technical features and the algorithm features that are interrelated and work together with the technical features.
The necessary technical parameters include:
- the computer hardware and its connection;
- the complete steps of the computer algorithm;
- the data required by the algorithm (eg, training data and sampling data); and
- the correlation between the parameters in the algorithm and various data in the application scenario.
If the above points can be conducted in the drafting phase of the description, the problem of insufficient disclosure will be largely avoided.
The problem of insufficient disclosure is different from other defects (eg, novelty, inventiveness and eligibility). Once the problem of insufficient disclosure is challenged in the examination process, the amendment to the application document (especially the amendment to the technical solution in the description) is limited due to the strict amendment requirement of the Chinese Patent Law. At that time, the applicant will face an embarrassing situation where the defect of insufficient disclosure cannot be overcome by the amendment.
Novelty and inventiveness (Articles 22.2 and 22.3)
An AI-related claim usually includes both technical and non-technical features (algorithms). The examination requirements on novelty and inventiveness of the AI-related invention can be roughly summarised into two points:
- overall judgement; and
- correlation effect.
For a solution that includes both technical and non-technical features, the non-technical features that are interrelated and work together with the technical features should be combined with the technical features as a whole to judge whether the claimed invention has novelty relative to a reference document.
The algorithm features or business rules and the method features, which functionally support and interact with the technical features, should be considered as a whole with the technical features.
The stipulation further emphasises the overall consideration of the technical and non-technical features, which highlights that the contribution of the non-technical features in solving the technical problem is accepted in the examination process. Therefore, in the drafting and office action handling phases, the applicant or attorney should pay attention to the integrity between the technical features and the non-technical features (algorithms) under the specific application scenario, and clearly emphasise the technical effect brought by the overall technical solution in the drafting phase, which will be conducive to overcoming the defects of lacking novelty or inventiveness.
The correlation between the technical and non-technical features is taken into consideration. This is also a consideration of the correlation between the AI algorithm and the specific application scenario. The technical problem (technical contribution) solved after the correlation between the AI algorithm (non-technical features) and the technical elements (technical features) in the specific application scenario will help to improve the persuasive power of the argument in the examination process of inventiveness.
At the drafting phase, the applicant or attorney should pay attention to describing the correlation between the AI algorithm and technical elements in the specific application scenario (eg, the correlation between the technical parameters in the specific application scenario and the parameters in the AI algorithm) and the technical problem (achieved technical effect) directly solved after correlation.
Eligibility of AI-related inventions
Attention should be paid to the technicality of a solution and the correlation between technical and non-technical features.
Sufficient disclosure of AI-related inventions
Attention should be paid to a specific application scenario of the non-technical features (algorithm features) and the disclosure of essential technical parameters.
Novelty and inventiveness of AI-related inventions
Whether in the drafting phase or in the office action handling phase, attention should be paid to the adaption of the principles of the overall judgement and correlation between the technical and non-technical features.