China National Intellectual Property Administration (CNIPA) has released the latest version of the drafted Amendment to the Examination Guidelines in mid-November of 2019. The Amendment was formulated to introduce a new chapter governing the rising inventions involving algorithms or business rules and methods, in response to the growing demands for IP protection over artificial intelligence, block chains, and business rules and methods. This is the first draft, which was made public a few weeks ago to solicit comments from the IP community and all those interested. Although a further modified version differing from this draft is likely to be issued, the draft may nonetheless serve as a basis to forecast a finalized working framework about China’s patent examination over AI and other modern technologies.

The first perspective in the new chapter denotes that features of algorithm or business rules shall not be separated from the technical features in order to be eligible for a patent. The Guidelines remind examiners to consider comprehensively all features recited in a claim so as to determine the traditional “3Ts,” which establishes an invention for patent shall propose a technical means to resolve a technical problem by yielding a technical result. Due to the large quantities of applications handled daily, each examiner spends very limited time on one application. Consequently, an examiner may easily issue a rejection after simply identifying algorithms as an element in the body of a claim. By having this reminder to the examiner, it is believable that the probability of rather arbitrary rejections will reduce optimistically.

Patent Law in Article 25 sets forth six non-patentable subject matters which are consisted as the negative list. Mental rules and activities are one category of the six. The new chapter advises that a claim having technical features shall not be rejected as a mere mental activity. However, an invention is not patentable if, for instance, it is a mathematical modeling process based on algorithms with no technical features, or an entirely business rule such as a rebate method dependent on online consumers’ purchasing behaviors.

After clearing the negative list, the examiner shall subsequently investigate whether a claimed subject matter meets the definition for an invention patent. To carry this out, the examiner shall take into consideration each and every cited element in a claim. To be more specific, when a set of data computed by an algorithm has a definite technical character and concurrently the execution of said algorithm can directly present a course of a natural law to resolve a specific technical problem by producing technical effects, it suffices the requirement as an invention patent. One good example is a convolutional neural network (CNN) model training method which produces a CNN model able to identify random-sized graphics, which is a beneficial technical result. It is critical to note that each step in the algorithm shall be technicality-related.

Next, each and every feature recited in a claim, including the algorithm or business rules, shall all be carefully examined in the phase to determine novelty and inventive step. The examiner shall consider the technical features and the algorithm or business rule that are functionally relying on or interactively relating to said technical features in its entirety. It is deemed to have inventive steps when the algorithm or business rules “closely integrates with the technical features to collaboratively yield a technical solution” so as to resolve a technical problem and hereby produce technical results.

Lastly, how the algorithms or business rules are functionally relying on or interactively relating to the technical features, including their beneficial effects shall be clearly described in the specification. Such beneficial effects can be any improvements in quality, accuracy, precision, efficiency, system property, user’s experience, and the like. It is suggested that the technical “benefits” of the claimed invention must be indispensable in the disclosure, rather than simply being “different”.

Other than the above general concepts, the new chapter provides 10 examples establishing allowable independent claims for AI, business rules, and block chains, as well as rejected claims which contrastingly incorporate mental activities. Accordingly, they serves as a clearer guidance; especially the 10 examples which are very valuable and important for drafting an AI claim. By following the new Chapter, an applicant may therefore be able to better draft allowable claims in China.