Overview & Analysis
The Measures for Security Assessment of Outbound Data Transfers establish the most stringent of China's three cross-border data transfer pathways — the mandatory government security assessment — applicable to important data and high-volume personal information flows. Under these Measures, data processors must first conduct a risk self-assessment, then apply to the provincial-level cyberspace administration to submit materials to the national CAC for official review. The assessment evaluates national security, public interest, and personal information risks in detail, covering the legality and necessity of the transfer, the data protection level of the overseas recipient's jurisdiction, the scale and sensitivity of the data, the adequacy of contractual protections, and the recipient's technical and management capabilities. Assessment results are valid for two years; re-application is required if circumstances change materially within the validity period, and renewal must be initiated at least 60 working days before expiry.
This policy is particularly relevant for AI, as AI projects typically require large-scale cross-border data flows — especially for model training, inference, and data sharing. When AI projects involve the transfer of personal or sensitive data at volumes triggering the security assessment thresholds, companies must comply with the policy's requirements to ensure that the data is not exposed or misused during transit. By implementing security assessments and risk self-assessments, companies can mitigate compliance risks and provide a clear compliance path for global AI collaboration, ensuring data flows meet China's legal framework while avoiding the high risks associated with data breaches.
These Measures govern the security assessment pathway — the mandatory government review tier for the highest-volume or highest-sensitivity transfers. Understanding all three pathways is essential for route selection:
| Pathway | When Required / Available | Key Trigger Thresholds |
|---|---|---|
| Security Assessment (this document) | Mandatory — CII operators; high-volume PI; any important data | Any important data; or CII operators providing PI; or ≥1M PI processors; or cumulative ≥100K PI / ≥10K sensitive PI since prior Jan 1 |
| Certification | Non-CII; mid-range volumes; no important data | >100K but <1M individuals' PI (non-sensitive); or <10K sensitive PI (cumulative in calendar year) |
| Standard Contract | Non-CII; lower volumes | <100K individuals' PI; or <10K sensitive PI (cumulative in calendar year) |
In the AI context, this policy applies when AI projects involve important data or high-volume personal information flows. If an AI project involves the cross-border transfer of personal information, sensitive data, or critical information infrastructure data at volumes triggering the thresholds, it must undergo security assessment before the transfer can proceed.
1. Global AI Model Training Requiring Large-Scale Data Transfer
AI training often requires large-scale global data, which may involve personal or sensitive data. Under this policy, cross-border flows that reach the volume thresholds must undergo security assessments evaluating potential risks to national security or personal information rights. Companies must evaluate whether training data volumes trigger the cumulative threshold and whether any important data (which has no volume threshold) is involved.
2. Using Multinational Cloud Platforms and Computing Resources
AI projects relying on multinational cloud platforms or computing resources often involve cross-border data transfers. Where data volumes or types trigger the assessment thresholds, companies must conduct risk self-assessments and apply for official data security assessments to ensure appropriate data protection measures are in place. The overseas cloud provider's jurisdiction and data protection framework will be examined as part of the assessment.
3. Cross-Border Customer Data Sharing and AI-Driven Analytics
For AI projects involving customer information and business data, data transfer security is critical. Where cumulative cross-border customer personal information volumes since the prior January 1 reach 100,000 individuals, the security assessment pathway is required. Companies must assess data sensitivity, scale, transfer purposes, and the security measures and capabilities of the receiving party as part of both the self-assessment and the official application.
4. Cross-Border HR Management and Employee Data
AI applications in cross-border recruitment, employee analysis, and performance management often require processing large amounts of personal data. If cumulative cross-border employee personal information flows reach the threshold or involve sensitive personal information categories (health, biometrics, financial), the security assessment pathway applies. Companies must ensure that personal information and sensitive data are adequately protected and that contractual protections meet the mandatory requirements in Article 9.
5. Cross-Border AI R&D Collaboration Involving Sensitive Data
Cross-border collaborations between multinational companies in AI R&D often require sharing sensitive data and technologies. When the data involved includes important data — which has no volume threshold and always requires security assessment — or when personal information volumes reach the trigger levels, companies must conduct assessments. All partners must commit to complying with data protection regulations, and onward transfer restrictions must be explicitly contractualized.
These Measures provide a necessary compliance framework for cross-border data transfers in AI projects. Multinational companies must consider the compliance of their data flows from the project planning stage, especially when personal or sensitive data is involved. The standardized security assessment process helps companies ensure data security while avoiding legal disputes.
Incorporate Data Security Assessment into the AI Project Planning Stage
Cross-border data flow compliance should be addressed early in the AI project planning stage — especially when large-scale cross-border data transfers are involved. Require teams to review data flow paths at intake, determining which data requires a security assessment, which can use certification or standard contracts, and which is exempt. The 45-working-day assessment timeline means that waiting until a project is ready to launch before starting an application will cause significant delays.
Conduct Comprehensive Risk Self-Assessments Before Applying
The self-assessment is a prerequisite for application, not a formality. Require teams to assess data sensitivity, scale, transfer purposes, and the security measures and capabilities of the receiving party. The assessment must include a thorough analysis of potential risks — tampering, loss, leakage — and whether channels for individuals to protect their personal information rights are available. The self-assessment report is submitted as part of the application materials and must be accurate; false materials result in assessment failure and legal liability.
Draft Legal Documents with Mandatory Article 9 Clauses Before Applying
The legal documents (contracts) between the data processor and the overseas recipient are a required application submission and are directly assessed for adequacy. Article 9 specifies six mandatory minimum elements: transfer purpose/scope/method; overseas storage duration and post-expiry handling; onward transfer restrictions; security measures on control changes or force majeure; remedies, liability and dispute resolution; and emergency response for data incidents. These clauses must be in place before the application is filed — they cannot be drafted after assessment approval.
Track Cumulative Cross-Border Volumes Across All Project Streams
The threshold in Article 4(3) is measured cumulatively since January 1 of the previous year — not the current year. This means that even a data flow that seems small in isolation may trigger the threshold when combined with other cross-border transfers by the same data processor. Multinationals with multiple AI initiatives in China must track aggregate outbound data volumes across all projects, not just individually. A data flow tracking register is essential for threshold monitoring.
Establish Re-Assessment Monitoring and Renewal Processes
Assessment approval is valid for only two years, and re-assessment is required not only at expiry but also when circumstances change materially — including changes to transfer purpose or method, changes to the overseas recipient's jurisdiction's data protection environment, changes in the recipient's actual control, force majeure events, or amendments to legal documents. Build a compliance calendar that monitors all these triggers and ensures renewal applications are filed at least 60 working days before expiry — not just at the expiry date.
For multinational companies, these Measures provide a clearer and more systematic compliance framework for cross-border data flows. Through well-designed compliance pathways and risk assessments, companies can ensure that their AI projects meet China's legal requirements. Managers should integrate these compliance requirements into the day-to-day operations of AI projects to ensure that cross-border data flows are compliant, fast, and efficient.
Complete Regulatory Text
Article Index
- Articles 1–3 — Purpose, Scope & Governing Principles
- Articles 4–6 — Triggering Thresholds, Self-Assessment Obligations & Application Materials
- Articles 7–12 — Assessment Process, Review Criteria, Contract Requirements & Timeline
- Articles 13–17 — Re-assessment, Disagreement, Validity, Confidentiality & Revocation
- Articles 18–20 — Liability, Definition of Important Data & Effective Date
(1) The data processor provides important data overseas (no volume threshold — always triggers assessment);
(2) A critical information infrastructure operator, or a data processor processing personal information of more than one million individuals, provides personal information overseas;
(3) Since January 1 of the previous year, a data processor has cumulatively provided personal information of 100,000 individuals or sensitive personal information of 10,000 individuals overseas;
(4) Other circumstances as prescribed by the national cyberspace administration requiring the declaration of a security assessment for outbound data transfers.
(1) The legality, legitimacy, and necessity of the purpose, scope, and method of outbound data transfer and processing by the overseas recipient;
(2) The scale, scope, type, and sensitivity of the outbound data, and the risks that such transfer may pose to national security, public interests, or the lawful rights and interests of individuals or organizations;
(3) Whether the responsibilities and obligations undertaken by the overseas recipient, as well as its management and technical measures and capabilities, can ensure the security of outbound data;
(4) The risks of tampering, destruction, leakage, loss, transfer, or illegal acquisition or use of data during and after the outbound transfer, and whether channels for safeguarding personal information rights are smooth;
(5) Whether contracts or other legally binding documents related to outbound data transfer to be concluded with the overseas recipient (hereinafter collectively referred to as "legal documents") sufficiently stipulate data security protection obligations;
(6) Other matters that may affect the security of outbound data transfers.
(1) Application form;
(2) Outbound data risk self-assessment report;
(3) Legal documents to be concluded between the data processor and the overseas recipient;
(4) Other materials required for the security assessment.
The national cyberspace administration shall, within seven working days from receipt of the application materials, determine whether to accept the application and notify the data processor in writing.
(1) The legality, legitimacy, and necessity of the purpose, scope, and method of outbound data transfer;
(2) The impact of the data protection policies, laws, and cybersecurity environment of the country or region where the overseas recipient is located on the security of outbound data, and whether the data protection level of the overseas recipient meets the requirements of Chinese laws, administrative regulations, and mandatory national standards;
(3) The scale, scope, type, and sensitivity of the outbound data, and the risks of tampering, destruction, leakage, loss, transfer, or illegal acquisition or use during and after the transfer;
(4) Whether data security and personal information rights can be fully and effectively protected;
(5) Whether the legal documents to be concluded between the data processor and the overseas recipient sufficiently stipulate data security protection obligations;
(6) Compliance with Chinese laws, administrative regulations, and departmental rules;
(7) Other matters deemed necessary for assessment by the national cyberspace administration.
(1) The purpose, method, and scope of outbound data transfer, and the purpose and method of data processing by the overseas recipient;
(2) The location and duration of data storage overseas, and measures for handling outbound data after the storage period expires, the agreed purpose is fulfilled, or the legal documents are terminated;
(3) Binding requirements for onward transfer of outbound data by the overseas recipient to other organizations or individuals;
(4) Security measures to be taken when substantial changes occur in the actual control or business scope of the overseas recipient, or changes occur in the data protection policies, laws, or cybersecurity environment of its country or region, or in the event of force majeure affecting data security;
(5) Remedies, liabilities for breach, and dispute resolution mechanisms for violations of data security protection obligations under the legal documents;
(6) Requirements for emergency response in the event of risks such as tampering, destruction, leakage, loss, transfer, or illegal acquisition or use of outbound data, and mechanisms to safeguard individuals' personal information rights.
The data processor shall be responsible for the authenticity of submitted materials. Deliberate submission of false materials shall result in failure of the assessment and legal liability shall be pursued in accordance with the law.
The assessment results shall be notified to the data processor in writing.
(1) Changes in the purpose, method, scope, type of data transfer, or the purpose or method of processing by the overseas recipient that affect data security, or extension of overseas storage duration for personal information or important data;
(2) Changes in the data protection policies, laws, or cybersecurity environment of the overseas recipient's country or region, force majeure events, changes in actual control of the data processor or overseas recipient, or amendments to legal documents affecting data security;
(3) Other circumstances affecting outbound data security.
Upon expiration of the validity period, if outbound data transfer activities need to continue, the data processor shall reapply at least 60 working days before expiration.
数据出境安全评估办法
（2022年7月7日公布，自2022年9月1日起施行，第11号令）
来源：中国网信网
条文索引
（一）数据处理者向境外提供重要数据；
（二）关键信息基础设施运营者和处理100万人以上个人信息的数据处理者向境外提供个人信息；
（三）自上年1月1日起累计向境外提供10万人个人信息或者1万人敏感个人信息的数据处理者向境外提供个人信息；
（四）国家网信部门规定的其他需要申报数据出境安全评估的情形。
（一）数据出境和境外接收方处理数据的目的、范围、方式等的合法性、正当性、必要性；
（二）出境数据的规模、范围、种类、敏感程度，数据出境可能对国家安全、公共利益、个人或者组织合法权益带来的风险；
（三）境外接收方承诺承担的责任义务，以及履行责任义务的管理和技术措施、能力等能否保障出境数据的安全；
（四）数据出境中和出境后遭到篡改、破坏、泄露、丢失、转移或者被非法获取、非法利用等的风险，个人信息权益维护的渠道是否通畅等；
（五）与境外接收方拟订立的法律文件是否充分约定了数据安全保护责任义务；
（六）其他可能影响数据出境安全的事项。
（一）申报书；
（二）数据出境风险自评估报告；
（三）数据处理者与境外接收方拟订立的法律文件；
（四）安全评估工作需要的其他材料。
国家网信部门应当自收到申报材料之日起7个工作日内，确定是否受理并书面通知数据处理者。
（一）数据出境的目的、范围、方式等的合法性、正当性、必要性；
（二）境外接收方所在国家或者地区的数据安全保护政策法规和网络安全环境对出境数据安全的影响；境外接收方的数据保护水平是否达到中华人民共和国法律、行政法规的规定和强制性国家标准的要求；
（三）出境数据的规模、范围、种类、敏感程度，出境中和出境后遭到篡改、破坏、泄露、丢失、转移或者被非法获取、非法利用等的风险；
（四）数据安全和个人信息权益是否能够得到充分有效保障；
（五）数据处理者与境外接收方拟订立的法律文件中是否充分约定了数据安全保护责任义务；
（六）遵守中国法律、行政法规、部门规章情况；
（七）国家网信部门认为需要评估的其他事项。
（一）数据出境的目的、方式和数据范围，境外接收方处理数据的用途、方式等；
（二）数据在境外保存地点、期限，以及达到保存期限、完成约定目的或者法律文件终止后出境数据的处理措施；
（三）对于境外接收方将出境数据再转移给其他组织、个人的约束性要求；
（四）境外接收方在实际控制权或者经营范围发生实质性变化，或者所在国家、地区数据安全保护政策法规和网络安全环境发生变化以及发生其他不可抗力情形导致难以保障数据安全时，应当采取的安全措施；
（五）违反法律文件约定的数据安全保护义务的补救措施、违约责任和争议解决方式；
（六）出境数据遭到篡改、破坏、泄露、丢失、转移或者被非法获取、非法利用等风险时，妥善开展应急处置的要求和保障个人维护其个人信息权益的途径和方式。
数据处理者对所提交材料的真实性负责，故意提交虚假材料的，按照评估不通过处理，并依法追究相应法律责任。
评估结果应当书面通知数据处理者。
（一）向境外提供数据的目的、方式、范围、种类和境外接收方处理数据的用途、方式发生变化影响出境数据安全的，或者延长个人信息和重要数据境外保存期限的；
（二）境外接收方所在国家或者地区数据安全保护政策法规和网络安全环境发生变化以及发生其他不可抗力情形、数据处理者或者境外接收方实际控制权发生变化、数据处理者与境外接收方法律文件变更等影响出境数据安全的；
（三）出现影响出境数据安全的其他情形。
有效期届满，需要继续开展数据出境活动的，数据处理者应当在有效期届满60个工作日前重新申报评估。
