Differences in the Assessment of Inventiveness Between Chinese and Foreign Patents and Practical Insights for Cross-Border Patent Strategies

The “Decision of the National Intellectual Property Administration on Amending the Patent Examination Guidelines” (Administrative Order No. 84) was published on November 10, 2025, and took effect on January 1, 2026. This amendment refines the rules for assessing patent inventiveness, clarifying that technical features must be appropriately related and aligned with the technical problem they aim to solve. Furthermore, algorithmic features can only be considered as contributing to inventiveness if they are deeply integrated with hardware and industrial applications. This revision further highlights the institutional differences in inventiveness examination between China and other countries. Taking typical cases from the China National Intellectual Property Administration as a starting point, this article analyzes the practical implications of these differences.

I. Typical Case: Determination of Inventiveness in AI Training Patents

Case No. 7 of the China National Intellectual Property Administration’s 2025 Review and Invalidation Typical Cases (Case No.: 4W118569): The patent in question relates to technology for generating dynamic images of digital avatars driven by audio. By combining a single reference image with audio input, the patent enables the creation of personalized, dynamic virtual images. The invalidation requester argued that both the underlying neural network and the 3D facial reconstruction technology used in the patent were already known in the prior art, representing merely the application of general models to specific scenarios and thus lacking inventive step. Relying on the spirit of the newly revised regulations, the panel clarified that the inventive step of AI patents must be assessed in each of three stages: model design, model training, and model application, in order to define the boundaries of innovation. After careful review, the panel determined that the core innovation of the patent was not the generic algorithmic architecture itself, but rather a specialized fine-tuning training mechanism tailored specifically to the target individual. This mechanism works by comparing the differences between the reference image and the predicted image, calibrating expression and motion coefficients, and iteratively optimizing model parameters to achieve customized generation results. This particular training approach had not been disclosed in the prior art nor suggested by any existing industry technologies, and it demonstrably achieved technical effects such as cost reduction, efficiency improvement, and enhanced fitting accuracy. Consequently, the patent was upheld as valid.

II. Core Differences in inventive steps/obviousness Judgment Between China and Foreign Countries

Taking this case as an example, we can clearly observe the structural differences in inventive steps/obviousness  judgment between China and foreign countries:

China adopts a refined three-step approach to rigorously verify the closed-loop nature of technological contributions, requiring that technical problems, technical solutions, and technical outcomes be fully aligned. The 2025 edition of the “Patent Examination Guidelines” explicitly excludes conventional adaptation features that do not yield substantial improvements.

The European Patent Office (“EPO”) adopts a problem-solution approach. For hybrid inventions, it is necessary to distinguish between technical features and non-technical features; non-functional modifications may not be counted as contributing to inventive step. Decision G 1/19 of the EPO Boards of Appeal (Pedestrian Simulation) also clarifies that relying solely on the technical principles underlying the simulated object is insufficient to produce technical effects beyond those inherent in the computer implementation itself and thus does not constitute an inventive contribution.

The U.S. adopts the four-factor framework from the Graham case; following the KSR decision, the flexibility of patentability reviews has significantly increased. The USPTO’s examination guidelines list seven different approaches for determining obviousness, and technology solutions that are well-suited to specific scenarios tend to be more readily accepted. Nevertheless, examiners are still required to provide sufficient reasoning and may not make conclusions arbitrarily.

III. Practical Strategies for Cross-Border Patent Portfolio Development

In response to the above-mentioned differences, patent filing strategies worldwide must adopt a differentiated drafting approach:

• Chinese applications must rigorously establish a closed-loop relationship among technical issues, technical means, and technical effects. Emphasis should be placed on clearly outlining the model fine-tuning logic, proprietary training steps, and quantified technical outcomes, thereby mitigating the risk that creativity may not be recognized due to overly generalized descriptions.
• For foreign (non Chinese) applications, you may appropriately emphasize the differentiated description of application scenarios, focusing on the overall differentiation of the solution and its industrial application value, thereby aligning with the generally apparent review standards in the U.S. and Europe.
• Avoid directly translating foreign patent texts for Chinese applications; instead, prepare documents specifically tailored to the review logic of both domestic and overseas jurisdictions.

The 2025 Edition of the Examination Guidelines marks the entry of China’s patent inventive step examination into a stage oriented toward substantive contribution, further solidifying the differentiation between China’s examination system and those of Europe and the U.S. Only by adopting a dual-track approach—building strategies that cater to both Chinese and foreign legal systems and adapting to a variety of the examination rules in different jurisdictions—can enterprises take the initiative in cross-border intellectual property battles.

References:

[1] National Intellectual Property Administration. Guidelines for Patent Examination (Administrative Order No. 84) [S]. 2025.

[2] CNIPA. Typical Cases of Patent Reexamination and Invalidation in 2025 (No. 7, Case No. 4W118569) [EB/OL].

[3] EPO Enlarged Board of Appeal. G 1/19 Pedestrian Simulation [EB/OL]. 2021.

[4] EPO. Case Law of the Boards of Appeal 2025[R]. Part I, Section D.

[5] USPTO. Manual of Patent Examining Procedure (MPEP) §2143[S]. [6] KSR International Co. v. Teleflex Inc., 550 U.S. 398(2007).