Artificial intelligence (AI) technologies have become deeply integrated into the intellectual property field and are now important tools for helping IP professionals improve the quality and efficiency of their work. A clear understanding of the relevant laws, regulations, and regulatory requirements is essential not only for ensuring the lawful operation of service providers, but also for frontline practitioners to manage professional risks, safeguard the legitimate rights and interests of clients, and uphold professional credibility.

I. Core Laws, Regulations, and Regulatory Requirements for AI Compliance

At present, AI compliance in China follows a framework combining “general regulation + sector-specific regulation.” Intellectual property professionals should pay particular attention to the following requirements.

(I) Core Laws and Regulations Generally Applicable to AI 

1. The Cybersecurity Law, Data Security Law, and Personal Information Protection Law together form the core framework for AI data compliance. Key compliance requirements include:

① Data processing must be subject to security management throughout the entire lifecycle, with data classified and protected according to its category and level, so as to prevent leakage, tampering, destruction, and misuse.

② Processing of personal information must follow the principles of lawfulness, legitimacy, necessity, and good faith; separate consent is required for the processing of sensitive personal information.

③ Data circulation and transfer, including the provision, sharing, and transmission of data to external parties, must comply with applicable laws and regulations.

④ Confidentiality must be strictly maintained for non-public intellectual property information, including trade secrets, unpublished technical solutions, and pending IP application materials, in order to prevent unauthorized disclosure or improper use for model training.

2. The Provisions on the Administration of Deep Synthesis Internet Information Services are the first dedicated departmental regulations governing the application of deep synthesis technologies. They cover six major categories of technological scenarios, including facial and voice generation and image enhancement. 

Key requirements include:

① Generated content must be appropriately labelled;

② Real-name authentication and content management requirements apply, and the creation or dissemination of unlawful content and false news is strictly prohibited;

③ Separate consent must be obtained for the editing of biometric information;

④ Service providers must establish management systems covering the entire service process.

3. The Interim Measures for the Management of Generative Artificial Intelligence Services  clarify the primary responsibilities of providers of generative AI services. They apply to services provided to the public within China that generate text, images, audio, video, and other content. Internally developed applications that are not made available to the public within China are not subject to these measures.

Whether AI services are developed in-house or procured from external providers, commercial secrets and personal information must be protected. Algorithms, data, platforms, and other advantages must not be used to engage in unfair competition. Review and audit records should also be maintained to satisfy regulatory traceability requirements and evidentiary needs in the event of disputes.

4. The Measures for the Labelling of Artificial Intelligence-Generated and Synthesized Content require online information service providers to apply two types of labels to all AI-generated or AI-synthesized content: explicit labels that are clearly perceptible to users, and implicit labels embedded in the file metadata.

Concealing the involvement of AI in content creation, or deleting or tampering with such labels to circumvent disclosure obligations, may expose the relevant party to regulatory scrutiny and compliance risks.

(II) AI Regulatory Requirements Specific to the Intellectual Property Sector

1. The revised Patent Examination Guidelines (effective from 1 January 2026) introduce additional examination rules relating to artificial intelligence and big data.

Key practical points include:

① Verify the identity of the inventor and prevent AI-generated fictitious inventors;

② Patent applications involving AI models and training must clearly describe the necessary modules, layers or connection relationships, as well as the steps, parameters, and other elements required for training, to a standard sufficient to enable reproduction;

③ Patent applications involving algorithmic features will not be granted if they violate laws or social morality, or are detrimental to the public interest.

2. The Regulations on Patent Agency, the Measures for the Administration of Patent Agency, and the Provisions on the Supervision and Administration of Trademark Agency establish professional standards for the industry. Their core compliance principles are consistent with the principles underlying AI compliance, including good faith, quality control, prohibition of fraudulent or bad-faith applications, confidentiality obligations, and prohibition of improper or misleading publicity.

The Code of Professional Ethics and Practice Discipline for Patent Agents  (revised in 2025) expressly provides that: “Patent agents shall not use artificial intelligence to directly generate final patent application documents for filing.”

II. Key Practical Considerations, with Scenarios and Typical Cases

The use of AI in the intellectual property field is mainly concentrated in areas such as information retrieval and analysis, document preparation and processing, infringement monitoring, and client services. In practice, particular attention should be paid to the following three core areas.

(I) Selection of AI Tools and Compliance of Foundation Models

1. Preference should be given to AI tools and large language models that have completed the required regulatory filings and are covered by legitimate commercial licences. Particular attention should be paid to verifying the legality of their training data and ensuring that such data does not infringe the intellectual property rights or personal information rights of others. AI tools that lack the necessary qualifications, have unclear origins, or are overseas tools without adequate compliance safeguards should not be used.

[Case 1] In February 2026, the Economic Information Daily reported that some companies were offering software services that used AI to draft patent applications, allowing applicants to purchase the software and use AI to prepare patent filings. One such company in Anhui charged fees ranging from RMB 1,000 to RMB 8,800 . The company was subsequently summoned for an administrative regulatory meeting and ordered to improve its technical control measures within a prescribed period, with the aim of closing technical loopholes that could facilitate fraudulent patent filings.

2. Official risk warnings should be strictly observed: On 1 April 2026, the China National Intellectual Property Administration issued a warning concerning the risks of using certain AI agents to draft patent application documents. The first risk is leakage of technical information. The second is the risk of substantive deficiencies, including “AI hallucinations” that may result in logical inconsistencies, unclear descriptions of technical features, and other defects. The third is the risk of bad-faith or dishonest patent applications. Accordingly, applicants should strengthen their awareness of such risks and exercise caution when selecting compliant patent agency services. Patent agencies and patent agents should remain highly alert to the risks associated with the use of AI agents, refrain from using AI agents to engage in dishonest patent filing activities, and properly safeguard the legitimate rights and interests of their clients.

3. Enter into a formal service agreement with the AI tool provider: The agreement should specify responsibilities for data security, ownership of intellectual property rights, restrictions against using submitted data for model training, and compensation or indemnification for unauthorized disclosure or confidentiality breaches. High-risk arrangements such as informal cooperation or trial use without a written agreement should be avoided.

[Case 2] In March 2023, within 20 days after Samsung Electronics introduced an AI tool, three incidents involving the leakage of core data reportedly occurred. As Samsung had not entered into a formal enterprise-level service agreement with the provider of the AI tool, the parties had not agreed on restrictions concerning the reuse of submitted data. All input content was therefore reportedly incorporated by default into the provider’s model-training dataset and permanently stored on third-party servers. Following disclosure of the incidents, Samsung urgently introduced an internal ban on the use of the tool.

(II) Data Processing Compliance

Intellectual property professionals naturally handle substantial volumes of confidential information. The processing of data using AI must therefore strictly follow the principles of lawfulness, legitimacy, necessity, and minimum retention.

1. Data collection: When collecting corporate or personal information, if AI applications are involved, the purpose of use, such as AI-assisted searching or proofreading, should be clearly disclosed and consent obtained where required. The purpose of use must not be concealed, and excessive data collection should be avoided.

2. Data storage and protection: An AI data storage management system should be established, together with security measures such as access controls. Data retention periods should be clearly defined and should not exceed what is necessary for business purposes. Regular data security reviews should also be conducted to prevent AI tools from unlawfully retaining or disclosing data.

[Case 3] In 2025, an AI agent platform provided by a well-known international technology company was reported to contain a vulnerability. An attacker only needed to insert hidden instructions into a shared file, and when an employee later used the AI tool to conduct a search, the AI system’s retrieval-augmented generation (RAG) function could automatically retrieve the malicious document and execute the hidden instructions. The technology company subsequently fixed the vulnerability, but the incident serves as a warning that, as AI becomes increasingly integrated into workplace systems, prompt-injection risks may become a routine threat. Businesses should therefore redefine trust boundaries for AI, monitor RAG workflows, and restrict data sources.

3. Data use and sharing: Data used by AI must be strictly limited to the scope of authorization. Client data must not be used for AI model training or shared with third parties without authorization. Where sharing is genuinely necessary, consent should be obtained again where required, and a data security and confidentiality agreement should be entered into.

4. Data destruction: Once a matter has concluded, data that no longer needs to be retained should be promptly destroyed, and records of the destruction process should be maintained.

[Case 4] An AI image restoration tool was reported to contain security vulnerabilities. Although the application claimed that it would “not store user data,” it was in fact retaining users’ uploaded photographs and videos for extended periods in unencrypted cloud storage. In addition, the developer had hard-coded cloud access tokens into the application, meaning that anyone with basic technical knowledge could potentially gain access.  Key lesson: When businesses use AI to process data, they should ensure that the AI provider complies with the principle of data minimization.

(III) Compliance of AI-Generated Content

1. Comply with requirements for labelling AI-generated content: When publishing AI-generated content online, the use of AI should be proactively disclosed and appropriately labelled. For proposals, reports, and other materials submitted to clients or used internally, AI-related labels must not be maliciously deleted, altered, falsified, or concealed.

[Case 5] (2024) Zhe 0108 Min Chu No. 10311 (（2024）浙0108民初10311号)

A blogger, Li, published an AI-generated article entitled Is Alibaba Digital Holdings Limited Real? on an online account operated by Li. The article contained fabricated and inaccurate information, and its accompanying images included the “Alibaba” trade name and brand logo. Although Li marked the content as “AI-generated” in the backend of the account, no prominent label appeared on the user-facing version of the article, meaning that members of the public were unable to identify the special nature of the content’s source. The court held that the conduct constituted unfair competition.

2. Prevent intellectual property infringement: AI must not be used without authorization to imitate well-known works or reproduce copyrighted characters or elements. AI-generated designs should be subject to prior similarity searches. AI-generated content should also incorporate sufficient human creative input to enhance originality and reduce infringement risks.

[Case 6] Li extracted more than 20 images of the character “Medusa” from the animated series Battle Through the Heavens and used the “LoRA model training” function of an AI platform to train a model capable of generating images of the Medusa character for use by others. The court held that Li had infringed the rights of reproduction and communication through information networks, and awarded RMB 50,000 in economic losses and reasonable enforcement expenses.

3. Strengthen responsibility for human review: Documents and materials processed using AI must be reviewed and confirmed by a human to ensure that their contents are truthful, compliant, and accurate. Records of the review process should be maintained, with clear individual accountability. AI must not be relied upon to piece together technical solutions or generate fabricated application materials.

[Case 7] The Ningbo Intellectual Property Protection Center found that, in patent pre-examination matters handled by two agencies, the case materials contained AI-generated technical content. In accordance with the relevant rules, the Center suspended the patent application pre-examination services available to both agencies for one year.  

In addition, in a notice concerning typical quality issues identified in patent application pre-examination cases from January to March 2026, the Nanjing Intellectual Property Protection Center stated that four patent agencies were suspected of being involved in abnormal patent applications due to the inclusion of AI-generated technical content.

4. Clearly define ownership of intellectual property rights in AI-assisted outputs: In external agreements, ownership of intellectual property rights should be expressly stipulated based on the principle that the work is primarily human-led and involves substantive human intellectual input. Internally, ownership of employee-created work products should be clearly defined through internal policies.

[Case 8] (2024) Su 0581 Min Chu No. 6697 (（2024）苏0581民初6697号)

Lin used AI to create an image of a nighttime riverside scene, manually modified the image, and subsequently registered the copyright. Two companies later used the image without authorization, and the court found that their conduct constituted infringement. AI-assisted works may therefore still possess originality and qualify for copyright protection. Ownership should accordingly be agreed in advance to avoid future disputes.  

 

III. Conclusion

AI is profoundly reshaping working practices in the intellectual property industry, but improvements in efficiency must never come at the expense of compliance. From general data security requirements to sector-specific professional rules, and from the careful selection of AI tools to the allocation of rights and responsibilities in AI-generated content, every stage concerns the fundamental professional obligations of intellectual property service providers and the professional standing of individual practitioners. As the field of “AI + intellectual property” continues to evolve rapidly, sustainable and responsible use of the technology requires legal requirements to be embedded into operational procedures and risk awareness to become part of everyday professional practice. Compliance is not a constraint on development, but a foundation for the high-quality growth of the industry.

For business inquiries, please email: [email protected]

References

[1] Interim Measures for the Administration of Generative Artificial Intelligence Services. China Government Network. https://www.gov.cn/zhengce/zhengceku/202307/content_6891752.htm

[2] Notice of the All-China Patent Agents Association on the issuance of "Regulations on Professional Ethics and Practice Discipline for Patent Agents". All-China Patent Agents Association http://www.acpaa.cn/article/content/202509/6936/1.html

Can you "buy" a patent for just 600 yuan? - Uncovering the gray industrial chain of "patent fraud" Economic Information Daily https://www.jjckb.cn/20260205/707758071b9640f8985c65928b253cf9/c.html

[4] Cracking down on "unlicensed agents" and cleaning up "certificate hanging", Anhui takes strong measures to rectify the intellectual property agency industry. Wanmei Supervision. https://mp.weixin.qq.com/s/sYOiQLwgDVS2w2lO9Bc_oQ

[5] Risk Warning on Using Agents like OpenClaw to Write Patent Application Documents. China National Intellectual Property Administration. https://mp.weixin.qq.com/s/6ZqIze99Dx9veViu_HhnvQ

[6] Gemini Zero-click Vulnerability: Stealing Corporate Office Confidential Information Through RAG Data Poisoning. Kanxue Academy. https://www.secrss.com/articles/85933

[7]AI App Wondershare RepairIt Found Leaking User Images and Data. Amar