Recently, the Guidelines for Application of Artificial Intelligence Related Invention Patents (Draft for Comments) (hereinafter referred to as the Guidelines) were issued, aiming to solve the practical difficulties faced by AI technology in patent protection, and to promote the standardization and normalization of patent examination work.

1. Clarify the object requirements for patents of AI-related technologies.

The Guidelines clearly state that applications for AI-related technologies must be technical, capable of solving specific technical problems, and have technical feasibility.

2. Improve the adequacy of technical solution disclosure.

1) For algorithm logic and model parameter disclosures:

The applicant needs to clearly describe the operational logic and model design of the algorithm, such as the number of layers, connection methods, and training methods of the neural network. This can help the examiner quickly assess the inventiveness and practicality of the solution.

2) For transparency of data sources and technical paths:

The application documents shall clearly state the source, collection methods, and preprocessing methods of the dataset used, and disclose the steps and paths of technical implementation. Taking medical AI as an example, the applicant needs to demonstrate how the model completes lesion detection based on medical imaging data.

3) For integrity of experimental verification results:

Applicants are required to provide experimental verification data for the technical solution, including performance indicators, testing environment, and comparison results. This requirement helps to reduce patent invalidity caused by incomplete disclosure and enhance the stability of patent protection.

3. Emphasize the core role of technological contributions in inventiveness evaluation.

1) Focusing on technical effects:

Inventiveness no longer relies solely on the novelty of technical solutions, but is centered on whether they effectively solve existing technical problems. For example, the innovative contribution of a machine learning model lies in whether it can achieve higher accuracy or faster inference speed in image recognition tasks.

2) Encourage substantive innovation:

compared to formal parameter optimization, the Guidelines encourage applicants to demonstrate inventiveness in technical implementation methods or application scenarios. For example, a model for natural language processing significantly improves the accuracy of sentiment analysis through a novel word vector representation, meeting the requirements of inventiveness.

3) Cross disciplinary technology integration:

In interdisciplinary fields such as healthcare and transportation, inventiveness emphasizes substantive innovation in technology integration. This evaluation criterion motivates technology developers to break through the limitations of a single discipline and promote collaborative innovation across multiple fields.

4. Emphasize the protection rules for technical solutions in specific fields.

1) Combining industry standards:

For example, AI patents in the medical field need to clarify how the technical solution meets clinical standards and has practical technical effects on specific medical problems.

2) Balancing between the protection of innovative achievements and open collaboration:

The Guidelines not only protect technological achievements, but also prevent monopolistic technology protection through industry review standards, promoting technological openness and collaborative innovation.

5. Strengthen data and ethical compliance requirements.

1) Proof of data legality:

Especially in the fields of healthcare or personal privacy, applicants need to provide detailed explanations to prove that the data source is legal and complies with national laws and regulations.

2) Ethical review criteria:

The application documents must reflect whether the technical solution complies with ethical standards in data collection, processing, and use, especially in the processing of highly sensitive data such as genetic data and medical records.

3) Compliance ensures future development:

The strict requirements of data and ethics are not only technological limitations, but also the foundation of sustainable industry development, providing social recognition and legal protection for the large-scale promotion of artificial intelligence technology.

6. Clarify the patent protection boundaries of generative artificial intelligence technology.

1) Protecting technical solutions rather than generated content:

The output of generative AI, such as images or text, is not considered a patent object, but the technical implementation process of generating such content (such as model architecture and training methods) can be patented.

2) Avoiding the generalization of technological protection:

By defining generative AI, guiding the industry to focus on innovative technological methods, and preventing unfair competition caused by excessive protection of generated content.

7. Prevent duplication and low-quality in patent applications.

1) Raise the threshold for patent application:

Clear and detailed technical and inventiveness requirements oblige applicants to optimize the application content.

2) By optimizing the examination process and improving patent quality standards to guide industry resources be distributed to technologies with practical innovation value.

3) The high threshold application standards have restrained the unfair competition caused by low-quality patent applications.

4) The Guidelines not only provide clear guidance for patent applicants and examiners, enhancing feasibility for both patent applications and examinations, but also provide legal protection for the sustainable innovation and development of the AI industry.