A Comparative Study on the Examination Practice of AI-related Patent Applications from a Multination | Article

Bin CHEN, Senior Partner, Shanghai Beshining Law Office

Keywords: Eligibility, AI related inventions, Conflict Application

I. Introduction

The patent system is a crucial mechanism for stimulating scientific and technological innovation. With the rapid development of AI (Artificial Intelligence) technology, the number of AI-related inventions has been increasing. How to define the patentability of such inventions has become a core issue of concern globally. Patent offices in different jurisdictions show certain similarities and differences in the assessment of the patentability of AI-related inventions. This paper will explore the examination standards and practices of the IP5 (China, Europe, Japan, South Korea, and the United States) regarding the subject-matter eligibility, claim formats, novelty, and inventive step of AI-related inventions through a comparative study. Through the comparative analysis of the IP5, it reveals the different strategies and standards of the international patent system in the face of the challenges of emerging technologies, providing references for innovative entities and IP professionals in global portfolios in the AI-related fields.

II. Glossary

CNIPA：China National Intellectual Property Administration
Chinese Patent Law：Patent Law of the Peoples’s Republic of China
CNIPA Guidelines：Guidelines for Patent Examination 2023
USPTO：United States Patent and Trademark Office
35 U.S.C. ：United States Code Title 35 – Patents
MPEP：Manual of Patent Examining Procedure in U.S.
EPO：European Patent Office
EPC：European Patent Convention’
EPO Guidelines：Guidelines for Examination in the European Patent Office
JPO：Japan Patent Office
JP Patent Law：Japan Patent Act
JP Guidelines：Examination Guidelines for Patent and Utility Model in Japan
JP Examination Handbook：Examination Handbook for Patent and Utility Model in Japan
KIPO：Korean Intellectual Property Office
KR Patent Law：Korean Patent Act
KR Guidelines：Patent Examination Guidelines in Korea
KR AI Guidelines：Examination Guide in the Artificial Intelligence Field in Korea

III. Subject-matter Eligibility

Subject-matter eligibility is a basic requirement for an invention to be eligible for patent protection, and it is also the focus of this paper. Patent laws of various jurisdictions stipulate that only inventions meeting the requirements of subject-matter eligibility can be granted patent rights. AI-related inventions often involve computer programs/software, deep learning/neural network algorithms, models, etc. Therefore, subject-matter eligibility has always been the primary concern in our patent practice. Although the IP5 have their specific standards and procedures for determining the subject-matter eligibility of AI-related inventions, the determination mainly focuses on aspects such as whether it has technical characteristics, whether it is applied to a specific field, and whether it contains an innovative technical solution. The aim is to ensure that the patent system can not only encourage technological innovation but also avoid over-protecting non-patentable subject matters such as rules and methods for mental activities, and the laws of nature, etc.

III-1. CNIPA

Article 2 of the Chinese Patent Law defines the patentable subject-matter, and Paragraph 1 of Article 25 lists the subject-matter excluded from patent protection, clearly stipulating that rules and methods for mental activities are not patentable. Section 4.2, Chapter 1, Part II of the "CNIPA Examination Guidelines" further explains the rules and methods of intellectual activities and provides non - exhaustive exclusion examples. In addition, Chapter 9, Part II of the "CNIPA Guidelines" specifically provides "Several Provisions on the Examination of Invention Patent Applications Involving Computer Programs", and Section 6 therein particularly provides relevant provisions on the examination of invention patent applications containing algorithm features or business rule and method features.

According to the examination standards of CNIPA, for AI-related inventions, it is necessary to examine whether they belong to the technical solutions described in Paragraph 2, Article 2 of the Chinese Patent Law. That is, the three elements of a technical solution - technical problem, technical means, and technical effect - should be considered and judged as a whole. In other words, technical features and algorithm features should not be simply separated from each other. Instead, all the contents recited in the claim should be regarded as a whole, and an overall analysis should be carried out on the technical means involved, the technical problem to be solved, and the technical effect obtained. Furthermore, Section 6.1.2, Chapter 9, Part II of the "CNIPA Examination Guidelines" provides the examination criteria for various specific situations, and Examples 1 to 10 in Section 6.2 provide examination examples.

Briefly speaking, regarding eligibility, CNIPA first determines whether the claimed solution belongs to rules and methods for mental activities. If not, it further determines whether the three technical elements are adopted.

III-2. EPO

The EPC does not define the patentable subject-matter like Article 2 of the Chinese Patent Law. Instead, similar to Paragraph 1 of Article 25 of the Chinese Patent Law, 52(2) EPC lists non-patentable subject-matters, and 52(3) EPC stipulates that only when a European patent application or European patent relates to such subject-matter or activities as such should it be excluded.

EPO Guidelines G-I, 1 stipulates that an invention must have "technical character", that is, it must be related to a technical field, solve a technical problem, and have technical features. EPO Guidelines G-II, 2 provides a two-step approach for assessment of patentability. The first step is called patent eligibility step, indicating that one technical feature is sufficient to meet the eligibility requirement. That is, if the claimed solution involves or uses technical means, then the solution is patent-eligible. EPO Guidelines G-II, 3.3 provides more examples of the technical purposes that artificial intelligence and machine learning can be applied to.

Compared with CNIPA, in terms of eligibility judgment, EPO is more lenient, and only one technical means is sufficient.

III-3. JPO

Similar to the Chinese Patent Law, Paragraph 1 of Article 2 of the JP Patent Law defines an invention as "a highly advanced creation of technical ideas utilizing the laws of nature". Paragraph 3 (i) of Article 2 of the JP Patent Law stipulates that a computer program belongs to a "product", and Paragraph 4 of Article 2 of the JP Patent Law defines a computer program. Different from CNIPA and EPO, the JP Patent Law does not provide a list of excluded non-patentable subject-matters. Instead, it is stipulated in Section 2.1, Chapter 1, Part III of the JP Guidelines, which stipulates that a statutory "invention" must be "a creation of a technical idea utilizing the laws of nature". Section 2.2, Chapter 1, Part III of the JP Guidelines also specifically provides points to consider in examination on an invention utilizing computer software.

Regarding the eligibility of "software-related inventions", Section 2.1, Chapter 1, Appendix B of the JP Examination Handbook provides a specific process of determination and particular examples. First, it is determined whether the invention as a whole utilizes a law of nature, and thus is regarded as "a creation of a technical idea utilizing the laws of nature", that is, whether it constitutes a statutory "invention". If not, it does not meet the eligibility requirements. If it cannot be determined, it is necessary to determine whether the information processing by the software is concretely realized by using hardware resources, that is, it depends on whether a specific information processor or an operation method thereof depending on intended use is constructed through cooperation of the software and the hardware resources.

Compared with CNIPA, when determining eligibility, both CNIPA and JPO require considering the claimed solution as a whole. In addition, CNIPA further requires the three technical elements, while JPO is more lenient on the condition that it is constructed through cooperation of the software and the hardware resources, and there is no clear requirement for technical problems and technical effects.

III-4. KIPO

Similar to CNIPA and JPO, Paragraph 1 of Article 2 of the KR Patent Law defines an invention as "a highly advanced creation of technical ideas utilizing a law of nature". Similar to JPO, the KR Patent Law also does not provide a list of excluded non-patentable subject-matters, but stipulates such in Section 4.1, Chapter 1, Part III of the KR Guidelines. Section 3 of the KR AI Guidelines provides that the criteria on patent eligibility of an AI-related invention are the same as those for a computer software-related invention, and Section 3.1.2 provides a specific method of assessment. First, it is determined whether the claimed subject-matter is directed to a creation of technical ideas utilizing a law of nature" (including the principle of determining as a whole). If not, it does not meet the eligibility requirements. If it cannot be determined, it should be reviewed whether the invention belongs to the case where information processing by software is specifically implemented by means of hardware.

It can be seen that the processes of KIPO and JPO in determining eligibility are very similar. When determining eligibility, both require considering the claimed solution as a whole. In addition, CNIPA further requires the three technical elements, and similar to JPO, KIPO is more lenient that information processing by software is specifically implemented by means of hardware, and there is no clear requirement for technical problems and technical effects.

III-5. USPTO

35 U.S.C.§101 defines the patentable subject-matter as process, machine, manufacture, or composition of matter. MPEP § 2106 provides judicial exceptions for patent-eligible subject-matter, namely abstract ideas, laws of nature and natural phenomena. For patent subject matter eligibility, MPEP § 2106 I provides two criteria and a specific analysis flowchart (see MPEP § 2106 III). First, it is evaluated whether the subject-matter of the claim falls within one of the four statutory categories. If so, it may be necessary to further conduct the Alice/Mayo test, a two-part test to identify calims that are directed to a judicial exception (Step 2A) (MPEP § 2106.04). If so, it is evaluated whether the claim has additional elements that provide an inventive concept (also called "significantly more" than the recited judicial exception) (Step 2B) (MPEP § 2106.05). The evaluation of additional elements includes both individual evaluation and combined evaluation with other elements of the claim, that is, all elements of the claim are considered as a whole. Step 2B is usually considered to be to search for an inventive concept. For example, whether additional elements belong to well-known, conventional, or routine activities.

For Step 2A, MPEP § 2106.04 further provides a two-prong inquiry. First, it is determined whether the claim recites a judicial exception (Prong One). If so, it is further determined whether the recited judicial exception is integrated into a practical application of that exception. These two-prongs represents Step 2A in the Alice/Mayo test.

It can be seen that in the process of determining eligibility, the criteria adopted by USPTO are different from or not similar to those of the other four offices, and the entire determination process is more complex. In the evaluation, if the claimed subject-matter is directed to a judicial exception, it is first necessary to evaluate whether there is a practical application, and then evaluate whether there is any additional element that provides an inventive concept.

IV. Claim Formats

As AI-related inventions often involve programs or algorithms, the claim formats thereof are the same as those of the claims for computer programs related inventions.

IV-1. CNIPA

In CNIPA, the claim of computer program related invention can be written as a method claim or a product claim, such as a device for implementing the method, a computer-readable storage medium, or a computer program product (Section 5.2, Chapter 9, Part II of the CNIPA Guidelines).

For example, before CNIPA, the following formats can be adopted:

[Method Claim] (Claim 1): A method for..., comprising: step A; step B; and step C.

[Product Claim in the form of a Functional Module Architecture]: A device/apparatus/system for..., comprising: means for implementing step A; means for implementing step B; and means for implementing step C.

[Product Claim of a mixture of software and hardware]: A computer device/apparatus/system, comprising a memory in which a computer program is stored, and a processor, wherein the computer program, when executed by the processor, can implement the steps of the method of Claim 1.

[Computer-Readable Storage Medium Claim]: A computer-readable storage medium on which a computer program is stored, wherein the computer program is executable by a processor to implement the steps of the method of Claim 1.

[Computer Program Product Claim]: A computer program product, comprising a computer program, wherein the computer program is executable by a processor to implement the steps of the method of Claim 1.

IV-2. EPO

For computer-implemented inventions (CII), EPO Guidelines F-IV, 3.9.1 provides, in a non-exhaustive manner, different acceptable claim formulations, such as:

[Method Claim] (Claim 1):

A computer-implemented method, comprising steps A, B, ...

A method carried out by a computer, including steps A, B, ...

[Device/Apparatus/System Claim] (Claim 2):

A data processing device/apparatus/system, comprising means for carrying out the steps of the method of Claim 1.

A data processing device/apparatus/system, comprising means for carrying out step A, means for carrying out Step B, ...

A data processing device/apparatus/system, comprising a processor adapted to/configured to perform the steps of the method of Claim 1.

[Computer Program (Product)] (Claim 3):

A computer program (product), comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method of Claim 1.

A computer program (product), comprising instructions which, when the program is executed by a computer, cause the computer to carry out steps A, B, ...

[Computer-Readable (Storage) Medium/Data Carrier] (Claim 4):

A computer-readable (storage) medium, comprising instructions which, when executed by a computer, cause the computer to carry out the steps of the method of Claim 1.

A computer-readable (storage) medium, comprising instructions which, when executed by a computer, cause the computer to carry out steps A, B, ...

A computer-readable data carrier having stored thereon the computer program (product) of Claim 3.

A data carrier signal carrying the computer program (product) of Claim 3.

In addition, EPO Guidelines G-II, 3.6.3 also provides that a computer-implemented data structure or data format embodied on a medium or as an electromagnetic carrier has technical character as a whole and thus falls within the definition of an invention. That is, "a medium storing a data structure" or "an electromagnetic carrier embodying a data structure" is also an acceptable claim format.

Compared with CNIPA, in terms of acceptable claim formats, EPO is more extensive. Its data carriers comprise not only physical tangible carriers but also electromagnetic waves. The content carried by the carriers can be not only computer programs but also data structures.

IV-3. JPO

For software-related inventions, Section 1.2.1.1, Chapter 1, Appendix B of the JP Examination Handbook provides acceptable claim formats, including:

1. Invention of a method.

2. Invention of a product, including a program, structured data, , a data structure, and computer-readable storage media.

Example 1: A program for causing a computer to execute step A, step B, step C, ...

Example 2: A program for causing a computer to function as means A, means B, means C, ...

Example 3: A program for causing a computer to implement function A, function B, function C, ...

Example 4: Structured data including a data element A, a data element B, a data element C, ...

Example 5: A Data structure including a data element A, a data element B, a data element C, ...

Example 6: A computer readable recording medium which records a program for causing a computer to execute a process A, a process B, a process C, ...

Example 7: A computer readable recording medium which records a program for causing a computer to function as means A, means B, means C, ...

Example 8: A computer readable recording medium which records a program for causing a computer to implement a function A, a function B, a function C, ...

Example 9: A computer readable recording medium which records structured data including a data element A, a data element B, a data element C, ...

In addition, according to Section 1.2.1.2, Chapter 1, Appendix B of the JP Examination Handbook, if it is clear in consideration of description, drawings and the common general knowledge at the time of filing that the claimed invention is a "program", even though the claimed subject matter is any word other than the "program" (for example, "module", "library", "neural network", "support vector machine" or "model"), the claimed invention is handled as the "program".

That is to say, as long as it can be clearly determined whether the claim is an "invention of a method" or an "invention of a product", JPO is very open-minded about the claim format. Compared with CNIPA and EPO, in terms of acceptable claim formats, JPO can even cover data structure and structured data per se.

IV-4. KIPO

For computer-related inventions, Section 2.2.2.2 of the KR AI Guidelines provides acceptable claim formats, including:

1. Method invention.

2. Product invention, such as an apparatus specified with functions, a computer-program-readable medium, a computer program recorded in a medium, and a recording medium having data structure.

Example 1: A computer-readable medium having a program recorded thereon, wherein the program makes a computer execute step A, step B, step C, ...

Example 2: A computer-readable medium having a program recorded thereon, wherein the program makes a computer to operate as means A, means B, means C, ...

Example 3: A computer-readable medium having a program recorded thereon, wherein the program makes a computer implement function A, function B, function C, ...

Example 4: A computer program recorded in a medium to make a computer to execute step A, step B, step C, ...

Example 5: A computer-readable medium having data recorded thereon, wherein the data comprises structure A, structure B, structure C, ... and operates on a computer.

It should be particularly noted that computer program per se is not allowed.

Unlike CNIPA, EPO, and JPO, at KIPO, a computer program product per se is not allowed, and is shall be recorded in a medium. Similar to JPO and EPO, at KIPO, recording media with a data structure is also allowed. In addition, program signals/program signal sequence are not allowed at KIPO.

IV-5. USPTO

35 U.S.C.§101 lists four categories of statutory subject-matters, namely process, machine, manufacture, or composition of matter. According to MPEP § 2106 I, a process defines actions, and the other three categories define the types of physical or tangible "things" or "products". Non-limiting examples of claims that are not directed to any of the statutory categories include:

Products that do not have a physical or tangible form (without any structural recitations), such as information (often referred to as "data per se") or computer program per se (often referred to as "software per se");

Transitory forms of signal transmission (often referred to as "signals per se"), such as propagating electrical or electromagnetic signal or carrier wave; and

Subject matter that the statue expressly prohibits from being patented, such as humans per se.

It can be seen that among the five offices, the claim formats provided by USPTO are the most limited.

V. Novelty

The literal expressions for novelty in the five offices are basically the same. Briefly speaking, it means that an invention does not belong to the prior art and there is no conflicting application. However, the definitions and extents of the conflicting application in the five offices are slightly different.

V-1. CNIPA

Paragraph 2 of Article 22 of the Chinese Patent Law defines novelty, that is, the claimed technical solution does not belong to the prior art and does not fall within the scope of the contents recorded in the conflicting application. At CNIPA, a conflicting application does not belong to the prior art.

Section 3 of Chapter 3, Part II of the CNIPA Guidelines provides criterion for examination on novelty. In addition to applying the novelty standards used for general technical fields, Section 6.1.3 of Chapter 9, Part II of the CNIPA Guidelines further points out that when conducting a novelty examination on an invention patent application containing algorithm features or business rule and method features, all the features recorded in the claim should be considered, such features including both technical features, and algorithm features or business rule and method features.

That is to say, for AI-related inventions, CNIPA considers all the features in the claimed technical solution, including both technical features and non-technical features.

V-2. EPO

54 EPC provides the definition of novelty by European Patent Office. It is highly similar to the provision on novelty as stipulated in the Chinese Patent Law. The difference is that a conflicting application falls within the extent of the prior art (but can only be used to evaluate novelty).

EPO Guidelines G-VI provides the examination of novelty, and there are no special provisions for AI-related inventions. Paragraph I.C.1 of the 10th edition of the case law of the EPO Boards of Appeal[1] states that "the first step in deciding whether an invention is new is to define the prior art, the relevant part of that art, and the content of that relevant part; and the next step is to compare the invention with the prior art thus defined, and see whether the invention differs from it", and Paragraph I.C.4 states that "for an invention to lack novelty, its subject-matter must be clearly and directly derivable from the prior art, and all its features (not just the essential ones) must be known from the prior art".

That is to say, for AI-related inventions, EPO applies the novelty examination criteria for general technical fields. Similar to CNIPA, it considers all the features in the claimed solution.

V-3. JPO

Paragraph 1 of Article 29 of the JP Patent Law defines novelty by exclusion, and Article 29-2 gives the definition of a conflicting application. Literally, the provisions on novelty at JPO are similar to those at CNIPA and EPO, but in fact, there are differences in the determination of the prior art and conflicting application. Specifically, at JPO, the time boundary of the prior art is refined to a specific moment of time, not a specific date; and the conflicting application excludes patent applications or patents with the same inventor and the same applicant.

Section 3 of Chapter 2, Part III of the JP Guidelines provides the procedure of determining novelty. According to Section 2.2.1, Chapter 1, Appendix B of the JP Examination Handbook, for software-related inventions, the novelty examination needs to consider all the features in the claimed solution. In addition, Section 2.2.4, Chapter 1, Appendix B of the JP Examination Handbook also gives special points to note regarding determination of novelty of software-related inventions.

That is to say, for AI-related inventions, similar to CNIPA and EPO, JPO applies the novelty examination criteria for general technical fields and considers all the features in the claimed solution.

V-4. KIPO

Similar to JPO, Paragraph 1 of Article 29 of the KR Patent Law also defines novelty by exclusion, and Paragraph 3 gives the definition of a conflicting application. Similarly, similar to JPO, at KIPO, a conflicting application excludes patent applications or patents with the same inventor and the same applicant.

Chapter 2 of Part III of the KR Guidelines provides the procedure for novelty examination. According to Section 3.2 of the KR AI Guidelines, for AI-related inventions, the novelty examination needs to consider the claimed solution as a whole, that is, consider all of its features, rather than considering each individual feature separately.

That is to say, for AI-related inventions, similar to CNIPA, EPO, and JPO, KIPO applies the novelty examination criteria for general technical fields and considers all the features in the claimed solution.

V-5. USPTO

35 U.S.C.§102 defines novelty. Since the provisions of U.S.C. 102 involve some unique legal concepts, different from the "first-to-file" system adopted by the patent laws of the other four offices, the U.S. patent law before the AIA (Leahy-Smith America Invents Act) adopted the "first-to-invent" system; after the AIA, the U.S. patent law adopted the "first-inventor-to-file" system. This article does not elaborate on the novelty judgment criteria of the U.S. patent law but simply makes comparisons with the other four offices in some aspects for AI-related patent applications. At USPTO, 35 U.S.C. § 102(a) clarifies the definition of the prior art, which includes a concept similar to "conflicting application". Similar to EPO, at USPTO, the so-called "conflicting application" fall within the extent of the prior art. However, the difference is that it can also be used to evaluate inventive step. Similar to JPO and KIPO, at USPTO, the so-called "conflicting application" excludes applications from the same applicant. Moreover, at USPTO, "conflicting application" can also include published PCT applications designating the U.S.

MPEP § 2131 states that when evaluating novelty, the disclosure that is available as prior art must teach every element required by the claim under its broadest reasonable interpretation, that is, all the features of the claimed solution need to be considered.

For AI-related inventions, USPTO does not provide any specific evaluation criteria on novelty. Similar to CNIPA, EPO, and JPO, it applies the novelty examination criteria for general technical fields and considers all the features in the claimed solution.

VI. Inventive Step

Since AI-related inventions often involve programs or algorithms, and even include business rule and method features, the determinations of inventive step in different countries are not completely consistent.

VI-1. CNIPA

Paragraph 3 of Article 22 of the Chinese Patent Law defines inventiveness. Chapter 4 of Part II of the CNIPA Guidelines provides criteria for examination on inventiveness, that is, the three-step approach. For the examination of invention patent applications containing algorithm features or business rule and method features, Section 6.1.3 of Chapter 9, Part II of the CNIPA Guidelines provides the corresponding criteria for examination on inventiveness, which states "in examination on inventiveness for an invention patent application containing both technical features and algorithm features or business rule and method features, the algorithm features or business rule and method features that functionally support and interact with the technical features should be considered with the technical features as a whole. "Functionally support and interact with each other" means that the algorithm features or business rule and method features are closely integrated with the technical features, together form technical means to solve a technical problem and thus can achieve a corresponding technical effect."

That is to say, for AI-related inventions, in addition to applying the inventiveness examination criteria for general technical fields, CNIPA further requires that non-technical features that functionally support and interact with technical features should be considered with the technical features as a whole, and non-technical features shall not be separated.

VI-2. EPO

56 EPC defines inventive step by the European Patent Office. The EPO Guidelines G-VII provides the examination criteria for inventive step, that is, the "problem-solution approach", which is similar to the three-step approach of CNIPA. In addition, for claims comprising both technical and non-technical features (referred to as mixed-type inventions), the EPO Guidelines G-VII, 5.4 provides specific examination rules, which emphasize that when assessing the inventive step of such an invention, all the features which contribute to the technical character of the invention should be taken into account. The specific steps include: (i) determining the features which contribute to the technical character of the invention based on the technical effects achieved in the context of the invention; (ii) selecting the closest prior art based on the features identified in step (i) that contribute to the technical character of the invention; and (iii) identifying the differences between the closest prior art and the claimed invention, and determining the technical effect(s) of these differences, in the context of the claim as a whole, in order to identify from these differences the features which make a technical contribution and those which do not: (a) if there are no differences, a novelty objection is raised; (b) if the differences do not make any technical contribution, a lack of inventive step objection is raised; and (c) if the differences include features making a technical contribution, the following steps applies: formulating the objective technical problem on the basis of the technical effect(s) achieved by these features; and judging whether the claimed technical solution to the objective technical problem is obvious to the person skilled in the art.

That is to say, for AI-related inventions, similar to CNIPA, all the features in the claim should be considered. The difference is that before applying the examination rules for general technical fields, EPO needs to first determine which features in the claim belong to those that contribute to the technical character of the invention, then select the closest prior art and make comparison therewith, and finally determine which of the distinguishing features make a technical contribution.

VI-3. JPO

Paragraph 2 of Article 29 of the JP Patent Law defines the inventiveness. Section 3 of Chapter 2, Part III of the JP Guidelines provides the criteria for examination of inventiveness. For software-related inventions, Section 2.2.3, Chapter 1, Appendix B of the JP Examination Handbook provides criteria and specific examples for examination of inventiveness.

For AI-related inventions, the assessment of inventiveness applies those for software-related inventions. At JPO, technical features and non-technical features are not differentiated during the assessment.

VI-4. KIPO

Paragraph 2 of Article 29 of the KR Patent Law defines inventiveness. Chapter 3 of Part III of the KR Guidelines provides the criteria for examination of inventiveness. For AI-related inventions, Section 3.2 of the KR AI Guidelines provides criteria and specific examples for examination of inventiveness. Similar to JPO, technical features and non-technical features are not differentiated during the assessment.

VI-5. USPTO

35 U.S.C. § 103 defines inventiveness (non-obviousness). MPEP § 2141 provides the examination guidelines for determining non-obviousness, giving the specific principles of the law of obviousness mainly through case laws of U.S. courts at various levels. At USPTO, for the determination of non-obviousness of AI-related inventions, no special rules are provided, and those for general technical fields are applied.

Comparison Table of the IP5

Eligibility

CNIPA

Article 2 and Article 25.1 of the Chinese Patent Law; Section 4.2, Chapter 1, Part II, and Sections 6.1.2 and 6.2, Chapter 9, Part II of the CNIPA Guidelines
Technical problem, technical means, and technical effect

EPO

52(2) EPC, and 52(3) EPC; EPO Guidelines G-I, 1, G-II, 2, and G-II, 3.3
Any technical means.

JPO

Articles 2.1, 2.3, 2.4 of the JP Patent Law; Sections 2.1 and 2.2, Chapter 1, Part III of the JP Guidelines; Section 2.1, Chapter 1, Appendix B of the JP Examination Handbook
By cooperation of the software and the hardware resources

KIPO

Article 2.1 of the KR Patent Law; Section 4.1, Chapter 1, Part III of the KR Guidelines; Section 3 of the KR AI Guidelines
information processing by software is specifically implemented by means of hardware

USPTO

35 U.S.C. § 101；MPEP § 2106 I, III；MPEP § 2106.04；MPEP § 2106.05
Practical application of judicial exception + any additional element that provides an inventive concept

Claim Formats

CNIPA

Section 5.2, Chapter 9, Part II of the CNIPA Guidelines
Method; Device/Apparatus/System; Computer-readable storage medium (having stored thereon a computer program)Computer program product

EPO

EPO Guidelines F-IV, 3.9.1, and G-II 3.6.3
Method; Device/Apparatus/System; Computer-readable storage medium (having stored thereon a computer program); Computer program product; Computer-readable data carrier (having stored thereon a program or adata structure; Data carrier signal (carrying a program or a data structure)

JPO

Sections 1.2.1.1 and 1.2.1.2, Chapter 1, Appendix B of the JP Examination Handbook
Method; Apparatus/Device/System; Program; Computer readable recording medium (having recorded thereon a program / a data structure); Data structure / Structured data

KIPO

Section 2.2.2.2 of the KR AI Guidelines
Method; Apparatus/Device/System; Computer readable medium (having recorded thereon a program / a data structure); Computer program recorded in a medium

USPTO

35 U.S.C. § 101；MPEP § 2106 I
Method; Apparatus/Device/System; Computer-readable medium

Novelty

CNIPA

Article 22.2 of the Chinese Patent Law; Section 3, Chapter 3, Part II of the CNIPA Guidelines; Section 6.1.3, Chapter 9，Part II of the CNIPA Guidelines
Having conflicting application; All the features shall be taken into account

EPO

54 EPC; EPO Guidelines G-VI; Paragraphs I.C.1 and I.C.4 of the 10th edition of the case law of the EPO Boards of Appeal
Having conflicting application; All the features shall be taken into account

JPO

Article 29 of the JP Patent Law; Section 3, Chapter 2, Part III of the JP Guidelines; Sections 2.2.1 and 2.2.4, Chapter 1, Appendix B of the JP Examination Handbook
Having conflicting application (different inventor or different applicant); All the features shall be taken into account

KIPO

Article 29 of the KR Patent Law; Section 2, Chapter 2, Part III of the KR Guidelines; Section 3.2 of the KR AI Guidelines
Having conflicting application (different inventor or different applicant); All the features shall be taken into account

USPTO

35 U.S.C. § 102；MPEP § 2131
Having concept similar to conflicting application (belonging to the prior art, not only being used for evaluating novelty, and different applicatn); All the features shall be taken into account

Inventive Step

CNIPA

Article 22.3 of the Chinese Patent Law; Chapter 4, Part II of the CNIPA Guidelines; Section 6.1.3, Chapter 9， Part II of the CNIPA Guidelines
non-technical features that functionally support and interact with technical features should be considered with the technical features as a whole

EPO

EPC 第56条；EPO Guidelines G-VII 5.4
all the features that contribute to the technical character of the invention should be considered

JPO

Item 2, Article 29 of the JP Patent Law; Section 3, Chapter 2, Part III of the JP Guidelines; Section 2.2.3, Chapter 1, Appendix B of the JP Examination Handbook
technical features and non-technical features are not differentiated

KIPO

Item 2, Article 29 of the KR Patent Law; Chapter 3, Part III of the KR Guidelines; Section 3.2 of the KR AI Guideline
technical features and non-technical features are not differentiated

USPTO

35 U.S.C. 103；MPEP § 2141
Examination rules for general technical fields are applied

VII. Comparative Analysis of Eligibility in a Hypothetical Case

Taking Example 39[2] provided by USPTO regarding subject matter eligibility as a hypothetical case, the following comparison is made based on the above-mentioned summary and analysis.

For reasons of economy, the background of this example case is omitted here (for details, please refer to the original text). The exemplary claim is as follows:

A computer-implemented method of training a neural network for facial detection comprising:

collecting a set of digital facial images from a database;

applying one or more transformations to each digital facial image including mirroring, rotating, smoothing, or contrast reduction to create a modified set of digital facial images;

creating a first training set comprising the collected set of digital facial images, the modified set of digital facial images, and a set of digital non-facial images;

training the neural network in a first stage using the first training set;

creating a second training set for a second stage of training comprising the first training set and digital non-facial images that are incorrectly detected as facial images after the first stage of training; and

training the neural network in a second stage using the second training set.

At CNIPA, it is necessary to determine whether the claimed solution adopts any technical means to solve a technical problem and achieve corresponding technical effect(s). The claimed solution relates to a computer-implemented method of training a neural network for facial detection. It clearly recites that the data processed in each step of the training method is image data and how to process the image data in each step, which reflects that the neural network training algorithm is closely related to image information processing. According to the background of this case, the problem to be solved by this solution is that the prior art does not provide a face detection model that can detect faces in distorted images while limiting the number of false positives. This problem is a technical problem. Using different training sets to train the neural network in different training stages is technical means in conformity with the laws of nature, and technical effects of being able to provide a robust face detection model that can detect faces in distorted images while limiting the number of false positives are achieved. Therefore, the claimed solution belongs to the technical solution specified in Paragraph 2, Article 2 of the Patent Law and thus is a patent eligible subject matter.

At EPO, it is necessary to determine whether the claimed solution adopts any technical means. The current claim is a computer-implemented method, so it meets the principle of “any technical means” and thus meets the patent eligibility requirement of EPO.

At JPO, it is necessary to determine whether the claimed solution utilizes the cooperation between software and hardware resources. As the claim clearly recites "computer-implemented", it is a specific operation method of an information processing device, which is established through the cooperation of software and hardware resources according to an intended use. Therefore, it meets the patent eligibility requirement.

At KIPO, it is necessary to determine whether the claimed solution is information processing by software that is specifically implemented by means of hardware. As the claimed solution is a computer-implemented method for training a neural network for facial detection, the computer, which is a hardware, specifically executes the software information processing program of neural network training. Therefore, it meets the patent eligibility requirement.

At USPTO, first, it is determined that the subject matter of the claim recites a method, and thus belongs to one of the four statutory categories. Then, it is necessary to further perform the Alice/Mayo test to determine whether the claim recites any of the judicial exceptions. In this example, the claim does not recite any mathematical relationships, formulas, or calculations. While some of the limitations may be based on mathematical concepts, the mathematical concepts are not recited in the claim. Further, the claim does not recite a mental process because the steps are practically performed in the human mind and cannot be performed by human. Finally, the claim does not recite any method for guiding, organizing, or arranging human activities. Thus, from the expression of the claim, it does not recite any judicial exception, so it meets the patent-eligibility requirement.

VIII. Conclusion

The five offices have different focuses when determining the patentability of AI related inventions. For example, regarding eligibility, CNIPA focuses on the integrity of the three technical elements, while EPO only requires the existence of technical means. JPO and KIPO require the cooperation or combination of software and hardware, and the determination of USPTO is the most complex. In terms of claim drafting, each country has specific regulations for AI related inventions, and the formats of protectable subject matters in USPTO are the most limited. For novelty, all of the five offices need to consider all the features in the claim, and they all have the concept of conflicting applications. However, the conflicting applications in EPO, though belonging to the prior art, cannot be used to evaluate inventive step. The conflicting applications in JPO and KIPO refer to different applicants or inventors. The conflicting application in USPTO belong to the prior art and can be used to evaluate inventive step. For inventive step, the five offices follow their respective examination principles and approaches. CNIPA emphasizes the overall consideration of technical and non-technical features, EPO focus on the features that contribute to the technical character, while JPO, KIPO, and USPTO apply the examination criteria for general technical fields in the examination of inventive step.

With the rapid development of AI technology globally, patent offices in different jurisdictions are constantly adjusting and improving their patent examination systems to adapt to the unique challenge of this emerging field. Each office is actively responding to the needs of intellectual property protection in the AI era, striving to find a balance between encouraging technological innovation and maintaining patent quality, and constructing a patent legal environment conducive to the research, development, and commercialization of AI technology. Through this article, innovative entities can have a clearer understanding of the examination practices of the five offices in patent applications for AI-related inventions, which is helpful for better global portfolio.

Reference:

[1] Case Law of the Boards of Appeal, Tenth Edition, July 2022. https://www.epo.org/en/legal/case-law/2022/index.html

[2] Subject Matter Eligibility Examples: Abstract Ideas. https://www.uspto.gov/sites/default/files/documents/101_examples_37to42_20190107.pdf