Selling company data to AI labs from China
Before company data goes to an AI lab, the privacy law where the people in it live asks one question: can anyone still be identified from it? This page explains the main laws in China, what counts as no longer personal, and what to check first. General information as of 2026-10-28, not legal advice: your counsel confirms what applies to your company.
At a glance
| Main laws | Three laws sit side by side. The Personal Information Protection Law (PIPL, in force since 1 November 2021) governs information about people. The Data Security Law covers all data by importance and sets a category called important data. The Cybersecurity Law adds duties for network operators and critical information infrastructure. Rules from the Cyberspace Administration of China (CAC) set how data leaves the country. |
|---|---|
| What it covers | The PIPL applies to handling personal information of people within mainland China, and also to handling outside China when the purpose is to provide products or services to people in China or to analyze their behaviour. Personal information is information that identifies or can identify a natural person, recorded electronically or otherwise, and it does not include anonymized information. |
| Roles | A personal information handler is an organization or person that decides on its own the purposes and methods of handling. Companies that license data about people in China are normally handlers. |
| Data outside the PIPL | Company records with no information about identifiable people are outside the PIPL, but the Data Security Law can still apply, and it treats some data as important data with restrictions on export. |
The main laws, and what each means for selling data
Personal Information Protection Law (PIPL)
Handling of personal information. It sets principles of legality, necessity and honesty, requires a legal basis for each handling, and gives individuals rights over their information.
Licensing records that identify people is a new handling and a provision of that information to another party. It needs a legal basis, and where the recipient is abroad, the extra steps for cross-border provision apply. Removing personal information first, to the PIPL's anonymization standard, takes the data outside the law.
Data Security Law
All data processing in China. The state sets up a categorized and classified system based on how important the data is and how much harm its loss or misuse would cause. Export of important data is regulated (Articles 21 and 31).
Even records without personal information can be caught if they are important data in a regulated field. Unless a department or region has notified data as important, a handler does not need to report it as important data for the export security assessment.
Provisions on Promoting and Regulating Cross-Border Data Flows (2024)
When data leaving China needs a security assessment, a standard contract with the recipient, or certification, and when it is exempt. They took effect on 22 March 2024.
Exemptions apply to data without personal information or important data produced in activities such as international trade, academic cooperation and manufacturing, and to small volumes: a handler that is not a critical infrastructure operator and has provided fewer than 100,000 people's personal information (not sensitive) overseas this year. Larger volumes need a standard contract or certification, and the largest need a security assessment.
Source: Cross-border provisions, English text (China Law Translate), and the CAC Chinese text
How China decides data is no longer personal
The PIPL draws a line that most laws blur. De-identification means handling personal information so that a specific person cannot be identified without additional information. Anonymization means handling it so that a specific person cannot be identified and the information cannot be restored.
The difference matters. Anonymized information is not personal information under Article 4, so it is outside the law. De-identified information is still personal information and stays covered. A set where names are replaced by stand-ins but a key still exists is de-identified, not anonymized.
Meeting the anonymization standard is hard to show, because it requires that the data cannot be restored. Counsel and an expert review decide whether a given set reaches it.
Sources: PIPL, Articles 4 and 73
Records that carry extra rules
Sensitive personal information
Information that could easily harm dignity, persons or property if leaked or misused, such as biometric identifiers, religious belief, specific identities, medical and health information, financial accounts and location tracking, and the information of minors under 14. It may be handled only for a specified purpose when fully necessary and with strict protection, and with the individual's independent consent.
Important data
A category under the Data Security Law for data whose loss would harm national security, the economy or public interests. Rules for exporting it are set by the Data Security Law and the cross-border provisions, and an assessment is needed for important data that has been notified or published as such.
Critical information infrastructure
Operators of critical information infrastructure face stricter export rules: personal information or important data they provide abroad needs a security assessment.
Sending data to an AI lab in another country
Where a handler truly needs to provide personal information abroad for business reasons, the PIPL requires one of the following: a security assessment organized by the cyberspace authority, certification by a professional body, a standard contract with the overseas recipient, or another condition in law. The handler must also tell the individuals who the recipient is, the purpose and the type of data, and obtain their independent consent (Articles 38 and 39).
The 2024 provisions set the thresholds. Counted from 1 January of the year, a handler that is not a critical infrastructure operator needs a security assessment for important data, for personal information of more than 1 million people, or for sensitive personal information of more than 10,000 people. It needs a standard contract or certification for personal information of 100,000 to 1 million people, or sensitive information of up to 10,000 people, and nothing of the kind below those levels. Critical information infrastructure operators need an assessment for any personal information or important data.
Exemptions cover data collected abroad and processed in China without bringing in domestic personal information, data from trade, transport, academic and similar activity that has no personal information or important data, and transfers needed to perform a contract with the individual. Assessment results last three years and can be extended.
Sources: PIPL, Articles 38 and 39, Provisions on Promoting and Regulating Cross-Border Data Flows, Articles 2 to 9, CAC, Chinese text of the provisions
Who enforces it
Questions about selling data from China
Is it legal to sell company data to an AI company from China?
It depends on what is in the data. Business records with no personal information or important data face fewer limits. Records about identifiable people need a legal basis, and sending them abroad triggers the cross-border steps. Counsel confirms the route.
What is the difference between de-identified and anonymized data in China?
De-identified data cannot identify a person without additional information, and it is still personal information. Anonymized data cannot identify anyone and cannot be restored, and it is outside the PIPL. The legal consequences differ.
Do I always need a security assessment to send data abroad?
No. Many cases are exempt or need only a standard contract or certification, depending on the kind of data and the volume in a year. Assessment is for important data, large volumes of personal information and critical infrastructure operators.
What about data about people outside China held by a Chinese company?
Article 4 of the cross-border provisions exempts personal information collected abroad and processed in China, if no domestic personal information or important data was added. The data may still be covered by the law of the place where the people are.
Where do I check the current rules?
Rules in this area change often. The English translations here are unofficial and the Chinese text is authoritative, so check the CAC site and take advice from counsel in China before sending data.
Didn't find your question? Ask us directly.
What we do about it
We find and remove the personal information in your company data, test whether anyone could still be identified, and write down how it was done for your counsel. If you license the data through us, the AI lab usually pays for that work, and the AI lab pays our broker fee, only after the deal closes. See how de-identification works and how licensing your data works.
Rules in other regions
- United States: HIPAA, California and other state laws, the FTC, and the records that carry extra rules.
- Canada: PIPEDA, Quebec Law 25, the provincial acts, and health information laws.
- United Kingdom: The UK GDPR, the Data Protection Act 2018, the 2025 Act and the ICO anonymisation guidance.
- European Union: The GDPR, the EDPB on anonymisation and AI models, and the EU AI Act.
- Australia: The Privacy Act 1988, the Australian Privacy Principles, OAIC guidance and health records laws.
- India: The DPDP Act 2023, the 2025 Rules, transfers abroad and the outsourcing exemption.