Overview & Analysis
The Cybersecurity Law of the People's Republic of China is the foundational statute of China's cybersecurity regime. It applies to the construction, operation, maintenance, and use of networks within China, as well as related supervision and administration. Its core purposes are to safeguard cybersecurity, protect cyberspace sovereignty, national security, and the public interest, protect the lawful rights and interests of individuals and organizations, and support the healthy development of informatization. Substantively, the law covers network operation security, critical information infrastructure protection, network information security, personal information handling, monitoring and early warning, emergency response, and legal liability. The 2025 amendment is especially important because it formally brings AI into the statute: Article 20 now expressly supports AI basic research, key technologies such as algorithms, training-data resources, and compute infrastructure, while also calling for stronger AI ethics, risk monitoring, assessment, and security supervision.
The law matters greatly for AI because it does not regulate only "traditional cybersecurity." It sets many of the baseline operating rules for AI in China — who counts as a network operator, what cybersecurity obligations apply, when logging, vulnerability remediation, incident reporting, real-identity requirements, and information-content controls matter, and how AI systems intersect with critical information infrastructure, important data, personal information, and cross-border transfers. Official commentary has expressly stated that the 2025 amendment brings AI into China's national cybersecurity legal system and strengthens coordination with the Data Security Law and the Personal Information Protection Law, as part of China's broader "development plus security" approach to AI.
The Cybersecurity Law is relevant very broadly — not only to foundation-model companies but to almost any entity that builds, operates, maintains, uses, or provides AI-related services through networks in China. After the 2025 amendment, AI is now expressly written into the base legal framework.
1. Developing, Deploying, or Operating AI Systems Through Networks
If an AI system runs over a network or is offered through one, the company will usually fall within the law's broad concept of a network operator and must perform baseline security obligations — building internal rules, assigning responsible personnel, adopting technical safeguards, monitoring system status and incidents, and keeping logs. Model services, AI assistants, SaaS AI tools, and internal intelligent platforms should be designed not only for functionality but also for compliance with underlying cyber-operation duties.
2. Processing Personal Information, User Data, or Training and Inference Data
The law requires network operators collecting and using personal information to follow legality, legitimacy, and necessity principles, disclose rules, explain purpose, method and scope, obtain consent, and take measures against leakage, damage, and loss. In AI practice, this directly affects prompts, chat histories, account data, behavioral data, training corpora, and inference logs — AI data governance must be integrated with personal-information compliance.
3. Content Generation, Publishing, Recommendation, or User Interaction
Network operators must strengthen management of information published by users and, upon discovering prohibited content, stop transmission, take disposal measures such as deletion, preserve records, and report to competent authorities. For generative AI, retrieval-augmented answers, recommender systems, and conversational products, content-governance obligations have to be built into product design and operations from the start.
4. AI Deployed in Critical Information Infrastructure or Key-Sector Environments
If AI systems are used in public communications, energy, transport, finance, public services, e-government, or similar important sectors, the law's heightened CII requirements become relevant — including dedicated security management, emergency drills, disaster recovery, procurement review, local storage, and annual assessment. High-impact AI is treated not merely as business innovation but as part of infrastructure security and national-security governance.
5. Cloud, Databases, Model Platforms, or Cross-Border Technical Architecture
The law requires network products and services to comply with mandatory national standards and prohibits malicious code, requiring prompt remediation, notice, and reporting of vulnerabilities with ongoing security maintenance. For AI projects, this affects vendor selection for models, cloud, databases, and middleware, as well as whether personal information and important data may be transferred overseas — CII operators must in principle store data domestically, with outbound transfer subject to security assessment.
For managers at multinational companies, AI compliance in China should not be understood only as a matter of privacy protection or content moderation. The Cybersecurity Law is a broader foundational framework covering network operations, product security, data handling, incident response, vendor management, and deployment in critical sectors.
Govern AI Projects First as Networked Systems
AI systems are often, first of all, networked systems. If a project runs in China over networks, connects to users, calls cloud resources, processes logs, or provides online services, it should be managed from the beginning through a network-operation security lens. A practical intake checklist should ask where the system is deployed, who owns security, how logs are retained, how vulnerabilities are remediated, and how incidents are escalated.
Build a Fast Tiered Governance Model
The way to balance speed and compliance is to stratify projects by risk. Low-risk internal productivity tools can go through lighter review. Projects processing moderately sensitive data or providing external-facing services need joint assessment by cybersecurity, data governance, and legal. Projects involving large volumes of personal information, important data, public interaction, key sectors, CII customers, or cross-border flows require enhanced review.
Make Logging, Monitoring & Incident Response Default Features
The law expressly requires monitoring and recording of operating status and cybersecurity incidents, keeping relevant logs for at least six months, and activating response plans upon incidents. For AI systems, model calls, permission changes, abnormal outputs, interface failures, data access, human overrides, and security incidents should all fall within the monitoring and audit trail system.
Put Security Maintenance Obligations into Vendor Contracts
The law requires network products and services to meet national standards, prohibits malicious code, and requires ongoing security maintenance. Procurement of AI models, cloud services, databases, middleware, and AI application platforms should specify vulnerability notification, patch timelines, logging support, incident cooperation, remote-access controls, maintenance periods, and exit arrangements.
Manage AI Data Governance Together with China Data Rules
The Cybersecurity Law already requires disclosure of collection and use rules, explanation of purpose/method/scope, consent, and prevention of personal-information leakage. The most effective approach is to require each important AI project to maintain a data map showing what is collected, whether it is used for training or inference, whether it is necessary for the service, whether it enters overseas systems, and how it can be deleted or corrected.
Design Content Governance Early for Public-Facing AI
The law requires network operators to stop transmission of prohibited information, delete it, preserve records, and report it. For generative AI, content governance must become a product capability, not merely a contract term. Management should ensure teams build prompt-risk controls, output interception, human escalation, complaint and reporting channels, record retention, and human review for higher-risk scenarios.
Apply a Higher Standard for CII-Adjacent Projects
Even where the company is not itself formally designated as a CII operator, if its AI system serves customers in finance, energy, transport, public services, e-government, or telecoms, assume a higher compliance standard — stronger access controls, disaster recovery, annual assessments, local deployment options, vendor confidentiality obligations, and procurement-review analysis. Manage these as a distinct class of high-impact projects from the outset.
Do Not Manage Projects with an "Old Law, Low Penalty" Mindset
The 2025 amendment increased the intensity of penalties for serious violations, especially where there are large-scale data leaks or partial or major loss of function in critical information infrastructure. AI security failures that might once have been viewed as ordinary IT issues — large log leaks, model-service failures disrupting key business, or poor vulnerability response — may now carry significantly more serious regulatory consequences in China.
The Cybersecurity Law is not a law that blocks AI development. It is the baseline governance boundary that makes larger-scale, longer-term AI deployment in China possible. The most effective approach is not to over-engineer compliance for every project, but to identify early which AI initiatives touch network operation security, content governance, personal information, critical sectors, or cross-border architecture — and then apply a higher level of control to those cases.
Complete Legislative Text
Table of Contents
Any individual or organization using a network shall abide by the Constitution and laws, observe public order, respect social morality, shall not endanger cybersecurity, and shall not use networks to engage in activities endangering national security, honor, and interests, inciting subversion of State power, inciting separatism or undermining national unity, promoting terrorism or extremism, promoting ethnic hatred or discrimination, disseminating violent, obscene, or pornographic information, fabricating or disseminating false information to disrupt economic or social order, or infringing upon the reputation, privacy, intellectual property, and other lawful rights and interests of others.
The State supports innovation in cybersecurity management methods and the use of new technologies such as artificial intelligence to improve the level of cybersecurity protection.
(1) formulate internal security management systems and operating procedures, determine the person responsible for cybersecurity, and implement cybersecurity protection responsibility;
(2) adopt technical measures to prevent acts endangering cybersecurity such as computer viruses, cyberattacks, and network intrusions;
(3) adopt technical measures to monitor and record network operating status and cybersecurity incidents, and retain relevant web logs for not less than six months as required;
(4) adopt measures such as data classification, backup of important data, and encryption;
(5) other obligations as prescribed by laws and administrative regulations.
Providers of network products and services shall continuously provide security maintenance for their products and services; they shall not terminate the provision of security maintenance within the period prescribed or agreed upon by the parties.
Where network products and services have functions for collecting user information, their providers shall expressly inform users and obtain their consent; where users' personal information is involved, they shall also comply with the provisions of this Law and relevant laws and administrative regulations on personal information protection.
The State implements a trusted online identity strategy, supports the research and development of secure and convenient electronic identity authentication technologies, and promotes mutual recognition among different electronic identity authentication systems.
(1) establish specialized security management bodies and security management personnel, and conduct security background checks on such personnel and personnel in key positions;
(2) regularly provide employees with cybersecurity education, technical training, and skills assessment;
(3) conduct disaster recovery backup for important systems and databases;
(4) formulate emergency response plans for cybersecurity incidents and conduct regular drills;
(5) other obligations as prescribed by laws and administrative regulations.
(1) conduct spot checks and testing of security risks of critical information infrastructure, put forward improvement measures, and where necessary may entrust cybersecurity service institutions to conduct inspections and assessments of network security risks;
(2) regularly organize operators of critical information infrastructure to carry out cybersecurity emergency drills, so as to improve their ability to respond to cybersecurity incidents and coordinate with one another;
(3) promote cybersecurity information sharing among relevant departments, operators of critical information infrastructure, relevant research institutions, and cybersecurity service institutions;
(4) provide technical support and assistance for emergency response to cybersecurity incidents and the restoration of network functions.
Where network operators process personal information, they shall comply with this Law and the provisions of laws and administrative regulations such as the Civil Code of the People's Republic of China and the Personal Information Protection Law of the People's Republic of China.
Network operators shall not collect personal information unrelated to the services they provide, shall not collect or use personal information in violation of laws, administrative regulations, or agreements between the parties, and shall process the personal information they preserve in accordance with the provisions of laws and administrative regulations and agreements with users.
Network operators shall adopt technical measures and other necessary measures to ensure the security of the personal information they collect and prevent information leakage, destruction, or loss. Where personal information leakage, destruction, or loss has occurred or may occur, they shall immediately take remedial measures, promptly inform users as required, and report to the relevant competent departments.
Providers of electronic information transmission services and application download services shall perform security management obligations; where they know that their users have engaged in the conduct prescribed in the preceding paragraph, they shall stop providing services, adopt disposal measures such as deletion, preserve relevant records, and report to the relevant competent departments.
Departments responsible for the security protection of critical information infrastructure shall formulate emergency plans for cybersecurity incidents in their respective industries and fields and regularly organize drills.
Emergency plans for cybersecurity incidents shall classify cybersecurity incidents according to factors such as the degree of harm and scope of impact after an incident occurs and shall prescribe corresponding emergency disposal measures.
(1) require relevant departments, institutions, and personnel to promptly collect and report relevant information and strengthen monitoring of cybersecurity risks;
(2) organize relevant departments, institutions, and professionals to analyze and assess cybersecurity risk information and predict the likelihood, scope of impact, and degree of harm of an incident;
(3) issue cybersecurity risk warnings to society and publish measures to avoid and mitigate harm.
Where an operator of critical information infrastructure fails to perform the cybersecurity protection obligations prescribed in Articles 35, 36, 38, and 40 of this Law, the relevant competent department shall order correction, issue a warning, and may impose a fine of not less than RMB 50,000 but not more than RMB 100,000; where correction is refused or consequences endangering cybersecurity are caused, a fine of not less than RMB 100,000 but not more than RMB 1,000,000 shall be imposed.
Where either of the acts described in the preceding two paragraphs causes serious cybersecurity consequences such as large-scale data leakage or partial loss of function of critical information infrastructure, a fine of not less than RMB 500,000 but not more than RMB 2,000,000 shall be imposed; where particularly serious cybersecurity consequences are caused, such as the loss of the principal functions of critical information infrastructure, a fine of not less than RMB 2,000,000 but not more than RMB 10,000,000 shall be imposed.
(1) installing malicious programs;
(2) failing to immediately take remedial measures for security defects, vulnerabilities, or other risks existing in its products or services, or failing to promptly inform users and report to the relevant competent departments as required;
(3) arbitrarily terminating the provision of security maintenance for its products or services.
Persons who have received public security administration penalties for violating Article 29 of this Law may not engage in work in cybersecurity management or key positions in network operation within five years; persons who have received criminal penalties may never engage in such work.
Where the act causes particularly serious impact or particularly serious consequences, the relevant competent department shall impose a fine of not less than RMB 2,000,000 but not more than RMB 10,000,000.
(1) refusing or obstructing supervision and inspection lawfully carried out by relevant departments;
(2) refusing to provide technical support and assistance to public security organs or state security organs.
(1) publishing or transmitting information prohibited from publication or transmission under Paragraph 2 of Article 13 of this Law and other laws and administrative regulations;
(2) infringing personal information rights and interests in violation of Paragraph 3 of Article 24 and Articles 43 through 45 of this Law;
(3) where an operator of critical information infrastructure stores personal information and important data overseas, or provides personal information and important data overseas, in violation of Article 39 of this Law.
Where staff members of the cyberspace administration department or relevant departments neglect their duties, abuse their powers, or engage in malpractice for personal gain and the act does not constitute a crime, they shall be given sanctions in accordance with the law.
Where a violation of this Law constitutes a violation of public security administration, a public security administration penalty shall be imposed in accordance with the law; where a crime is constituted, criminal liability shall be pursued in accordance with the law.
(1) "network" means a system composed of computers or other information terminals and related equipment that collects, stores, transmits, exchanges, and processes information according to certain rules and procedures;
(2) "cybersecurity" means the ability, through the adoption of necessary measures, to prevent attacks, intrusions, interference, destruction, and illegal use of networks, as well as accidental incidents, to maintain networks in a stable and reliable operating state, and to safeguard the integrity, confidentiality, and availability of network data;
(3) "network operator" means the owner, administrator, or network service provider of a network;
(4) "network data" means all kinds of electronic data collected, stored, transmitted, processed, and generated through networks;
(5) "personal information" means all kinds of information, recorded electronically or by other means, that can identify the identity of a natural person either alone or in combination with other information, including but not limited to the natural person's name, date of birth, identity document number, personal biometric information, address, telephone number, and so on.
中华人民共和国网络安全法
(2016年11月7日第十二届全国人民代表大会常务委员会第二十四次会议通过 根据2025年10月28日第十四届全国人民代表大会常务委员会第十八次会议《关于修改〈中华人民共和国网络安全法〉的决定》修正)
来源:中国人大网
目 录
任何个人和组织使用网络应当遵守宪法法律,遵守公共秩序,尊重社会公德,不得危害网络安全,不得利用网络从事危害国家安全、荣誉和利益,煽动颠覆国家政权、推翻社会主义制度,煽动分裂国家、破坏国家统一,宣扬恐怖主义、极端主义,宣扬民族仇恨、民族歧视,传播暴力、淫秽色情信息,编造、传播虚假信息扰乱经济秩序和社会秩序,以及侵害他人名誉、隐私、知识产权和其他合法权益等活动。
国家支持创新网络安全管理方式,运用人工智能等新技术,提升网络安全保护水平。
(一)制定内部安全管理制度和操作规程,确定网络安全负责人,落实网络安全保护责任;
(二)采取防范计算机病毒和网络攻击、网络侵入等危害网络安全行为的技术措施;
(三)采取监测、记录网络运行状态、网络安全事件的技术措施,并按照规定留存相关的网络日志不少于六个月;
(四)采取数据分类、重要数据备份和加密等措施;
(五)法律、行政法规规定的其他义务。
网络产品、服务的提供者应当为其产品、服务持续提供安全维护;在规定或者当事人约定的期限内,不得终止提供安全维护。
网络产品、服务具有收集用户信息功能的,其提供者应当向用户明示并取得同意;涉及用户个人信息的,还应当遵守本法和有关法律、行政法规关于个人信息保护的规定。
国家实施网络可信身份战略,支持研究开发安全、方便的电子身份认证技术,推动不同电子身份认证之间的互认。
(一)设置专门安全管理机构和安全管理负责人,并对该负责人和关键岗位的人员进行安全背景审查;
(二)定期对从业人员进行网络安全教育、技术培训和技能考核;
(三)对重要系统和数据库进行容灾备份;
(四)制定网络安全事件应急预案,并定期进行演练;
(五)法律、行政法规规定的其他义务。
(一)对关键信息基础设施的安全风险进行抽查检测,提出改进措施;
(二)定期组织关键信息基础设施的运营者进行网络安全应急演练;
(三)促进有关部门、关键信息基础设施的运营者以及有关研究机构、网络安全服务机构之间的网络安全信息共享;
(四)对网络安全事件的应急处置与网络功能的恢复等,提供技术支持和协助。
网络运营者处理个人信息,应当遵守本法和《中华人民共和国民法典》、《中华人民共和国个人信息保护法》等法律、行政法规的规定。
网络运营者不得收集与其提供的服务无关的个人信息,不得违反法律、行政法规的规定和双方的约定收集、使用个人信息。
网络运营者应当采取技术措施和其他必要措施,确保其收集的个人信息安全,防止信息泄露、毁损、丢失。
电子信息发送服务提供者和应用软件下载服务提供者,应当履行安全管理义务,知道其用户有前款规定行为的,应当停止提供服务,采取消除等处置措施,保存有关记录,并向有关主管部门报告。
负责关键信息基础设施安全保护工作的部门应当制定本行业、本领域的网络安全事件应急预案,并定期组织演练。
网络安全事件应急预案应当按照事件发生后的危害程度、影响范围等因素对网络安全事件进行分级,并规定相应的应急处置措施。
(一)要求有关部门、机构和人员及时收集、报告有关信息,加强对网络安全风险的监测;
(二)组织有关部门、机构和专业人员,对网络安全风险信息进行分析评估,预测事件发生的可能性、影响范围和危害程度;
(三)向社会发布网络安全风险预警,发布避免、减轻危害的措施。
关键信息基础设施的运营者不履行本法第三十五条、第三十六条、第三十八条、第四十条规定的网络安全保护义务的,由有关主管部门责令改正,给予警告,可以处五万元以上十万元以下罚款;拒不改正或者导致危害网络安全等后果的,处十万元以上一百万元以下罚款。
有前两款行为,造成大量数据泄露、关键信息基础设施丧失局部功能等严重危害网络安全后果的,由有关主管部门处五十万元以上二百万元以下罚款;造成关键信息基础设施丧失主要功能等特别严重危害网络安全后果的,处二百万元以上一千万元以下罚款。
(一)设置恶意程序的;
(二)对其产品、服务存在的安全缺陷、漏洞等风险未立即采取补救措施;
(三)擅自终止为其产品、服务提供安全维护的。
(一)网络,是指由计算机或者其他信息终端及相关设备组成的按照一定的规则和程序对信息进行收集、存储、传输、交换、处理的系统。
(二)网络安全,是指通过采取必要措施,防范对网络的攻击、侵入、干扰、破坏和非法使用以及意外事故,使网络处于稳定可靠运行的状态,以及保障网络数据的完整性、保密性、可用性的能力。
(三)网络运营者,是指网络的所有者、管理者和网络服务提供者。
(四)网络数据,是指通过网络收集、存储、传输、处理和产生的各种电子数据。
(五)个人信息,是指以电子或者其他方式记录的能够单独或者与其他信息结合识别自然人个人身份的各种信息,包括但不限于自然人的姓名、出生日期、身份证件号码、个人生物识别信息、住址、电话号码等。