Five key changes to the EPO Guidelines
We take a look at the five key changes to the European Patent Office’s Guidelines that came into force…
Read more
In January 2026, China’s patent regulatory framework has undergone substantial reform with amendments to the Patent Examination Guidelines which impact the examination of artificial intelligence (AI)‑implemented inventions. The changes respond to challenges created by rapid developments in machine learning, algorithmic decision-making, data‑dependent technologies, and industries that rely heavily on automated processes.
The revisions to the guidelines focus on compliance of AI-related inventions with laws and social morality, provide details on the requirements to sufficiently disclose the invention in the description and inventiveness.
The amendments to the guidelines explicitly allow examiners to consult the description when interpreting the scope of claims. The updated wording to second 6.1 now reads ‘Examination shall be conducted with respect to the solution for which protection is sought, namely the solution defined in the claims, and, where necessary, with respect to the contents of the description.’ As such, since the examiners will now be using the description to interpret the meaning of the claims, the updated guidelines have also provided clarification on the requirements for a sufficient disclosure of an AI-related invention.
According to the updated guidelines, to provide a sufficient disclosure of an AI-related invention, applicants should provide a clear description of model architecture including the hierarchical structure of the model, the modules and the layers as well as the specific training steps and key parameters. The description must also explain the technical relationship between the input data and the output data and a particular scenario in which the AI model can be implemented.
In addition, the updated guidelines also clarify the examination of inventive step for AI-related inventions. Previously, in the examination of inventive step, the examiners may have considered features of the invention disclosed in the description which do not form part of the claims. However, the updated guidelines require the features of the algorithm which contribute to the inventive step of the invention to be explicitly mentioned in the claims. This change ensures that claims reflect the technical contribution of the invention. It further requires applicants to anticipate examination challenges by including fallback options in the dependent claims. The shift signals CNIPA’s effort to improve claim clarity and examination efficiency.
Furthermore, mere application of an existing model to a new technical field without any making any improvements to the model or simply inputting a different dataset does not satisfy the inventive‑step requirement. Rather, applicants must demonstrate technical improvements to model structure, parameter selection, training mechanisms to establish an inventive step.
In China, a patent will not be granted for an invention that violates Chinese law, social morality or harm public interest. The updated guidelines now explicitly prohibit the grant of patents where algorithmic decision‑making or data‑collection methods breach the laws, social morality, or the public interest. The updated guidelines provide examples of violations which would result in the patent application being rejected. These examples include reliance on personal data collected without a person’s knowledge, such as facial‑recognition tools used without user consent and the emergency decision making models in autonomous vehicles making unethical decisions based on individuals characteristics such as age or gender.
The revision to the guidelines also clarifies that inventorship can only be attributed to a natural person and explicitly prohibits naming AI systems as inventors. Patent firms must now verify the authenticity of the identity information provided by the applicant to ensure that inventorship cannot be attributed to automated systems, even where AI tools contribute to the inventive process.
The recent amendments to China’s Patent Examination Guidelines represent a significant evolution in the treatment of AI‑implemented inventions. By providing these clarifications the reforms reflect China’s long‑term strategy for managing advanced technologies. The changes aim to provide the applicant with greater clarity of the expectations of the CNIPA with regards to AI-related patent applications.
If you have any questions about these amendments or for any other intellectual property query, please contact one of our attorneys.