Data, privacy and AI rules

Due in 60 days12Regulations190

12 due in 60 days · 190 regulations

China Generative AI Measures

In force China · In force Aug 15, 2023 · no upcoming deadlines

Official text
Add to calendar
Apple Calendar or OutlookGoogle Calendar Download .ics

Deadlines

Takes effectAug 15, 2023 · 3.2 years ago · Source
Generative AI measures take effectTraining data, labeling, content and user protection duties apply to public generative AI services.

Summaries for reference, not legal advice. Check the official text.

What it does

China's main rule for generative AI services offered to the public in China. Providers must use lawfully sourced training data, respect IP and personal information rights, label generated content, handle illegal content, and protect minors. Services with public opinion attributes or social mobilization capacity need a security assessment and algorithm filing.

Who it applies to
Organizations and individuals that provide generative AI services (text, image, audio, video) to the public in China, including via APIs. Internal R&D not offered to the public is excluded.
Penalties
Handled under the CSL, DSL, PIPL and Science and Technology Progress Law. Where no law applies, warnings, public criticism, rectification orders and suspension of service.
Enforced by
CAC with NDRC, MOE, MOST, MIIT, MPS and NRTA
Official name
Interim Measures for the Management of Generative Artificial Intelligence Services
Citation
CAC Order No. 15 (joint with NDRC, MOE, MOST, MIIT, MPS, NRTA)
Topics
AI, Online safety, Privacy
Verified 2026-09-25 gov.cn
Research notes

Adopted by CAC on 2023-05-23 and signed 2023-07-10 (the date used as enacted_date); published 2023-07-13. Security assessment and algorithm filing duties link to the Algorithmic Recommendation Provisions.

Related

AmendedAustralia
Australia Online Safety ActSets up the eSafety Commissioner and removal schemes for cyber-bullying of children, adult cyber-abuse, non-consensual intimate images and abhorrent violent material. It also sets Basic Online Safety Expectations with reporting duties, and binding industry codes and standards. The Unlawful Material codes and standards cover class 1A and 1B material such as child sexual exploitation and pro-terror content. The Age-Restricted Material codes cover class 1C and class 2 material such as online pornography, and require age assurance and other measures to protect children.In effect since Jan 23, 2022
AmendedAustralia
Australia Privacy ActAustralia's federal privacy law (13 Australian Privacy Principles, Notifiable Data Breaches scheme). The 2024 amendment adds a statutory tort for serious invasions of privacy, tiered civil penalties and infringement notices, a mandate for a Children's Online Privacy Code, clearer security duties, and privacy-policy transparency for substantially automated decisions from 10 December 2026. It also adds criminal doxxing offences.Next: Dec 10 Children's Online Privacy Code must be registered · in 44 days
In forceAustralia
Australia Social Media Minimum AgeRequires providers of age-restricted social media platforms to take reasonable steps to stop Australians under 16 from having accounts. Platforms may not rely only on government ID for age assurance, and must protect data collected for age checks and destroy it after use. eSafety has named platforms including Facebook, Instagram, Snapchat, TikTok, X, YouTube, Reddit, Threads, Twitch and Kick.In effect since Dec 10, 2025
In forceChina
China AI Content Labeling MeasuresRequires generative AI service providers to add explicit (visible) labels to AI-generated text, images, audio, video and virtual scenes, and implicit labels (metadata) to generated files. Content distribution platforms must detect and label AI-generated content, and app stores must check labeling functions when apps are listed. A mandatory national standard (GB 45438-2025) takes effect on the same date.In effect since Sep 1, 2025
In forceChina
China Algorithmic Recommendation ProvisionsRegulates recommendation, ranking, personalization, generation and dispatch algorithms used in internet services in China. Users must be able to switch off personalized recommendations and manage tags, and providers may not use algorithms for price discrimination or to induce addiction. Providers with public opinion attributes or social mobilization capacity must file their algorithms with the CAC within 10 working days of launching a service.In effect since Mar 1, 2022
In forceChina
China Anthropomorphic AI MeasuresRegulates AI companion and emotional interaction services that simulate human personality. Providers must tell users they are talking to AI, remind users after every 2 hours of use, intervene in crises such as self-harm, offer easy exit, and may not offer virtual partner or relative services to minors. Users' interaction data may not be shared or used for training on sensitive data without consent, and services with 1 million registered or 100,000 monthly active users need a security assessment filed with the provincial CAC.In effect since Jul 15, 2026
In forceChina
China Cross-Border Data Flow ProvisionsRelaxes China's data export regime by exempting common transfers (contracts with the individual, cross-border HR, emergencies, and non-CII transfers of PI of fewer than 100,000 people a year) and setting volume thresholds for when a CAC security assessment, standard contract or certification is needed. Data is not 'important data' unless regulators have notified or published it as such.In effect since Mar 22, 2024
In forceChina
China Cyber Incident Reporting MeasuresRequires network operators in China to grade cybersecurity incidents using an annexed grading guide and report incidents rated relatively major or above. CII operators must report within 1 hour, central government bodies within 2 hours, and other operators to the provincial CAC within 4 hours. A full post-incident report is due within 30 days after handling ends, and ransom demands must be disclosed.In effect since Nov 1, 2025
Questions about China Generative AI Measures
What are the China Generative AI Measures compliance deadlines?
Aug 15, 2023: Generative AI measures take effect.
When does China Generative AI Measures take effect?
China Generative AI Measures took effect on Aug 15, 2023.
Who does China Generative AI Measures apply to?
Organizations and individuals that provide generative AI services (text, image, audio, video) to the public in China, including via APIs. Internal R&D not offered to the public is excluded.
What are the penalties under China Generative AI Measures?
Handled under the CSL, DSL, PIPL and Science and Technology Progress Law. Where no law applies, warnings, public criticism, rectification orders and suspension of service.