On April 30, 2025, China National Intellectual Property Administration (CNIPA) issued a public notice seeking comments on the Draft Amendment to the Guidelines for Patent Examination (Request for Comments). This amendment focuses on improving the examination standards for patents in emerging fields and new business models, targeting issues in examination practice that are urgent and have generally reached consensus.

Key Points of the Draft Amendment:

1. Full disclosure of inventors' identity information required in the request form. Previously, only the identity information of the first inventor was required.

The revised provision states: "The inventor must be a natural person. The request form must include the identity information of all inventors. It is prohibited to list organizations, collectives, or artificial intelligence names, such as '×× Research Group' or 'AI ××'."

This requirement likely aims to reinforce that inventors must be natural persons, thus avoiding disputes regarding whether AI can possess the attributes of a living entity.

2. Patent agencies are responsible for the authenticity and validity of the identity information and contact details provided in the request form.

The amendment clarifies that the responsibility for verifying the authenticity of identity information lies with the agency submitting the request.

In the United States, Japan, and Europe, applicants themselves are responsible for the authenticity and validity of their identity and contact information, and agencies merely submit the information on their behalf. In practice, it is challenging for agencies to fully verify the information provided by applicants.

3. For same-day applications (where the same applicant files both an invention and a utility model application for the same invention on the same day), the determination of whether the applications relate to the same invention is based on a declaration in the request form. The applicant must waive the utility model right for the invention patent to be granted. The approach of modifying the invention application to distinguish it is no longer accepted.

[This amendment aims to conserve examination resources and reduce the burden on applicants. Applicants should carefully consider whether dual filing is necessary.]

Previously, when the same applicant filed both a utility model and an invention application for the same invention on the same date (referring only to the filing date), and the utility model right had not yet expired while a declaration was made at the time of filing, there were two ways to proceed: Modify the invention application; or waive the utility model right.

Under the amendment, the first option is no longer permitted. Hence, at the time of application, the applicant must clearly determine:

(1) Whether it is necessary for the utility model and invention to have consistent scopes of protection, i.e., whether to pursue dual filing;

(2) If the scopes are consistent, whether to proceed with a utility model or an invention patent.

4. Determination of Inventive Step (Inventiveness).

Technical features that do not contribute to the resolution of a technical problem generally do not enhance the inventiveness of the claimed invention.

[Applicants should not expect to obtain patents merely by stacking features to narrow the scope of claims. Descriptions of the technical effects of specific features and their synergy are becoming increasingly important.]

This amendment includes two main changes:

(1) Addition of the term “as a whole”: Inventiveness shall be assessed based on the technical solution as a whole defined by the claims, rather than evaluating the inventiveness of individual technical features;

(2) Deletion of the redundant sentence: “In addition, inventiveness should be evaluated based on the overall technical solution defined in the claims, not individual technical features.”

Addition of a clarification: “However, features that do not contribute to solving the technical problem, even if included in the claims, generally do not affect the assessment of inventiveness.”

Taken together, the first change maintains the principle of assessing the inventiveness of the technical solution. The deletion eliminates redundancy, while the new clarification imposes a further limitation on the overall assessment: in determining the scope of the “whole,” features that do not contribute to solving the technical problem are excluded from consideration.

5. Provisions Related to Applications Involving Artificial Intelligence, Big Data, etc.

(1) For patent applications related to artificial intelligence, big data, and similar technologies, it is clarified that the object of examination is the application documents.

(2) Addition of an examination based on Article 5, Paragraph 1 of the Patent Law: For invention patent applications that include algorithm features or features involving business rules and methods, if they contain content that violates laws, social ethics, or public interest — such as data collection, label management, rule-setting, or recommendation decisions that are unlawful, unfair or unjust, or exhibit discriminatory bias — such applications shall not be granted patent rights in accordance with Article 5, Paragraph 1 of the Patent Law.

[When drafting the specification, aspects such as data collection should be explained as needed — for example, stating that data collection occurs with user consent and complies with relevant personal privacy protection regulations.]

(3) Clarification of the disclosure sufficiency requirements for specifications and provision of relevant examples:

Since artificial intelligence algorithms or models may present a “black box” problem, the specification must disclose sufficient information to fulfill the requirement of sufficient disclosure.

Specifically: If the invention involves the construction or training of an AI model, the specification must clearly describe the essential modules, layers, or connection relationships of the model, as well as the specific training steps and parameters required. If the invention involves the application of an AI model or algorithm in a specific field or scenario, the specification must clearly explain how the model or algorithm is integrated with the specific domain or use case, including how the input and output data are defined, to demonstrate their intrinsic relationships.

The amendment also adds examination examples regarding whether the requirement for sufficient disclosure is met, such as:“A method for generating facial features”, and “A method for predicting cancer based on biological information”.

Amendments to Chapter 9:

Part II, Chapter 9

(1) The phrase “involving algorithm features or business rules and methods” is revised to “involving artificial intelligence, etc.”

In the context of artificial intelligence, the process of constructing models or algorithms may involve natural laws, such as the law of universal gravitation or the first and second laws of thermodynamics (hereinafter referred to as “laws”); it may also involve human-devised rules, such as in the newly added Example 2, where “the solution of the invention patent application is a method for constructing an emergency decision-making model for autonomous vehicles, which uses the pedestrian’s gender and age as obstacle data.” This constitutes a human-devised rule (hereinafter referred to as “rules”).

Therefore, unifying the original expression “involving algorithm features or business rules and methods” into “involving artificial intelligence, etc.” is more appropriate, as it encompasses both natural laws and human-devised rules. This revision eliminates the need to separate these elements into two categories for examination and facilitates a more consistent examination standard.

(2) In certain artificial intelligence models, the causal relationship between input and output information is not necessarily clear. For example, in newly added Example 18, which involves using deep learning for ship-counting image recognition, the AI system needs to be trained and learn from a “ship image dataset,” requiring a large volume of ship images for training. However, the exact process by which the model is trained, how the trained model operates, and the role each feature plays are not clearly understood—this means the model functions as a “black box.” Since the claims only describe the technical features that constitute the technical solution, the role of each feature may not be inferable. In such cases, it is necessary to refer to the description in the specification, including the disclosed technical means, the technical problem to be solved, and the technical effect achieved, to make an appropriate determination.

(3) Artificial intelligence systems are very likely to incorporate rules. For example, in newly added Example 1, the rule essentially constitutes a marketing rule: identity information is recognized based on the customer’s facial features to assist businesses in targeted marketing. The basis for such marketing is actually determined by human decisions—such as marketing based on a person’s height or their amount of assets. These human-devised rules often involve legal and ethical issues because they are, in essence, rules by which humans instruct machines to act through automated systems. Since laws and ethics are intended to regulate human behavior, actions carried out by machines under human direction also fall within the scope of legal and ethical governance.

In Example 1, “the use of image capture and facial recognition technologies in commercial locations such as shopping malls to conduct targeted mattress marketing clearly does not fall under activities necessary for maintaining public security. Moreover, the collection of customers’ facial information and the acquisition of their identity is done without their knowledge or consent.” Therefore, “this invention violates the law.”

It is also noteworthy that in this section, the term used is “invention patent applications containing algorithm features or business rules and methods,” rather than the newly introduced term “artificial intelligence.”

(4) In the drafting of patent applications, it is indeed possible that the technical solution contains features for which the causal relationships are unclear. This can be seen, for example, in the field of chemistry. Currently, many artificial intelligence models also exhibit such issues. For such patent applications, attention should be paid to clearly stating the input, model, and output. It is not necessary to explain causality, but as with chemical patents, it is advisable to provide detailed embodiments and descriptions of the technical effects. Reference may be made to newly added Example 21.

6. Newly Added: Examination Provisions for Invention Patent Applications Involving Bitstreams. Main content as follows:

(1) Examination of Subject Matter for Protection

Cases not eligible for patent protection:

If a claim is solely directed to a pure bitstream, it is not considered subject matter eligible for patent protection. Similarly, if all limiting elements of a claim (except for the claim title) relate only to a pure bitstream, the claim is also not considered eligible for protection.

Cases eligible for patent protection:

In the technical field of digital video encoding/decoding, if a specific video encoding/decoding method that generates the bitstream constitutes a technical solution in the sense of the Patent Law, then a method for storing or transmitting the bitstream based on that specific encoding/decoding method, as well as a computer-readable storage medium storing such a bitstream, may be considered patentable subject matter.

[The newly added provisions regarding bitstream-related invention patent applications clarify which types of bitstream claims are not allowable and which are allowable, including their drafting formats.]

(2) Specification Drafting Requirements

The specification shall fully disclose the specific video encoding/decoding method used to generate the bitstream.

If the invention relates to a method of storing or transmitting the bitstream, or to a computer-readable storage medium storing the bitstream, the specification shall also provide corresponding descriptions to support the claims.

[For method and medium claims involving bitstreams, the specification must contain sufficient supporting content.]

(3) Claim Drafting Requirements

Invention patent applications involving bitstreams generated by specific video encoding/decoding methods may be drafted as method, device, and computer-readable storage medium claims.

In a single invention patent application, it is generally required that claims related to storing, transmitting, and/or computer-readable storage media be based on the specific video encoding/decoding method claim, either by referring to it or by incorporating all its features.

[These provisions clarify how to draft various types of claims involving bitstreams.]

This revision regarding “bitstreams” may be an attempt to clarify the concept of “computer programs.”

Based on the given examples: A computer program essentially refers to an algorithm or model and may contain human-created inventive content. However, a “bitstream” is a set of rules defined by humans. This revision may be intended to distinguish between the two.

A signal (bit) is meaningless and only gains meaning through human convention. Therefore, this revision may emphasize that inventions based on the information content carried by the signal (i.e., human-defined meaning) are based on man-made rules and are not patentable subject matter. Only improvements beyond the signal—such as generation, storage, and transmission—are eligible for patent protection.

In simple terms, communication is vital to humanity. In communication, there are two parts: the carrier and the information content. For example, in a written letter, the characters are the carrier, while the meaning of the message is conveyed through those characters, which are governed by an agreed-upon set of rules (e.g., Chinese or English). Similarly, the bitstream is the carrier, and its meaning is presented via agreed-upon rules. This revision appears to be aimed at distinguishing human-defined rules from technical inventions.

Due to the potentially controversial nature of this revision, it is recommended to study it further once the final version is confirmed.

7. On Invalidation Requests

(1) If an invalidation request is determined not to represent the true intention of the requester, it shall not be accepted. This aims to prevent the filing of invalidation requests using another person’s name without authorization.

(2) Regarding the principle of “non bis in idem” (no repeated litigation) in invalidation proceedings: The phrase “same grounds and evidence” is commonly interpreted in theory and practice as meaning “identical or substantially identical grounds and evidence.” That is, if the grounds or evidence are merely adjusted or reformatted in appearance but are essentially the same in legal fact, they still fall within the scope regulated by the principle of non bis in idem.

(3) Newly added provision on “submission of amended texts”:

It is clarified that, in an invalidation proceeding, the patentee shall submit replacement pages for the full text and a comparison table of amendments.

Where multiple amended texts are submitted by the patentee during the same invalidation proceeding and all comply with Section 4.6.3, the revised rule specifies that the latest version submitted by the patentee shall be considered the official version for examination, and all earlier versions shall be deemed withdrawn.

This change aims to provide both parties with a clear expectation regarding which version will be examined.

8. On Patent Applications and Administrative Procedures

(1) Order of Examination

The section on “General Principles” adds:“At the applicant’s request, the patent application may be examined as needed, including prioritized examination, fast-track examination, or deferred examination.”

This reflects a multi-track examination system, aligning with practical needs of industries and applicants for demand-based review.

(2) Newly Added: “Fast-Track Examination”

It is clarified that patent applications that have passed pre-examination by a national intellectual property protection center or a fast-track rights protection center, and that meet the relevant procedures for fast-track examination, may undergo expedited review.

(3) On the Content of the Patent Certificate

For international applications or divisional applications, the names of the inventors/designers and applicants listed on the patent certificate refer to those at the time when the international application enters the national phase in China, or at the time the divisional application is filed.

(4) On Patent Term Adjustment

It is clarified that, in reexamination proceedings, even if the patent application documents are not amended, if the rejection decision is withdrawn based on new arguments or new evidence submitted by the reexamination requester, such a case constitutes a reasonable delay in the granting process.

“New arguments or new evidence” refer to those not submitted during the substantive examination stage—that is, arguments not raised or evidence not filed before the rejection decision was issued.

If the reexamination decision to overturn the rejection is made based on such newly raised arguments or evidence, the delay is considered reasonable. However, if the requester argues procedural violations during reexamination and the rejection is overturned solely due to procedural violations, such cases do not fall under the scope of “new arguments.”