EveryAILaw.comInterim Measures for Generative AI Services
Obligations Covered
Timeline
| Milestone | Date | Notes |
|---|---|---|
| Issued | Jul 10, 2023 | Date borne by CAC Order No. 15, issued with six other departments |
| Published | Jul 13, 2023 | Official CAC web publication |
| Effective | Aug 15, 2023 | Full application |
Related Regulations and Standards
Related instruments are selected from shared compliance obligations and jurisdiction coverage.
Generative AI Content Compliance
Requirements
| Requirement | Details |
|---|---|
| Content labeling | Label generated images, videos and other covered content under the Deep Synthesis Provisions (Art. 12); this article does not impose an all-output label |
| Content compliance | Providers and users must observe Article 4's content rules; providers bear online-information-content duties under Article 9 |
| Service agreement | Providers must agree service terms with users who register for their service, defining both sides' rights and duties (Art. 9); this is not a universal human-vs-AI interaction notice |
| Complaint mechanism | Providers must establish complaint and reporting channels, publish the process and feedback period, and handle complaints (Art. 15) |
Penalties
| Violation | Fine |
|---|---|
| Non-compliance | Article 21 applies penalties under relevant laws; absent a specified penalty there, authorities may warn, criticize and order correction, with suspension for refusal to correct or serious circumstances. Public-security and criminal consequences require their separate legal predicates |
Sources: CAC Official Text
Generative AI Risk Assessment
Requirements
| Requirement | Details |
|---|---|
| Security assessment and filing | Providers of services with public-opinion properties or social-mobilization capability must conduct a security assessment and complete algorithm filing under applicable rules (Art. 17) |
| Training data compliance | Providers must lawfully process training data, respect intellectual property and personal information rights, and take measures to improve data quality (Art. 7) |
| Data-labeling governance | Providers that label data during development must establish labeling rules, assess quality, sample-check accuracy and train and supervise labelers (Art. 8) |
| Cooperation with inspections | Providers must cooperate with lawful supervision and explain specified training-data and algorithm matters and provide needed support (Art. 19) |
Penalties
| Violation | Fine |
|---|---|
| Non-compliance | Article 21 applies relevant-law penalties or, where none is specified, warning, criticism and correction; suspension requires refusal to correct or serious circumstances. Criminal liability applies only when conduct constitutes a crime |
Sources: CAC Official Text
Generative AI Incident Response
Requirements
| Requirement | Details |
|---|---|
| Illegal content response | On discovering illegal content, take timely measures including stopping generation and transmission and eliminating the content as applicable, rectify through measures such as model optimization, and report to the competent departments |
| Unlawful user activity | On discovering unlawful use, take lawful and contractual measures such as warning, restricting functions, suspending or terminating service; preserve records and report to the competent departments |
Penalties
| Violation | Fine |
|---|---|
| Non-compliance | Article 21 applies relevant-law penalties or, where none is specified, warning, criticism and correction; suspension requires refusal to correct or serious circumstances. Criminal liability applies only when conduct constitutes a crime |
Sources: CAC Official Text
Cite this regulation
Permalink: https://everyailaw.com/regulation/cn-generative-ai/
JSON: https://everyailaw.com/api/v1/regulations.json
Attribution: EveryAILaw, PAICE.work PBC. “Interim Measures for Generative AI Services”, EveryAILaw.com, Jun 30, 2026. https://everyailaw.com/regulation/cn-generative-ai/
Agents: cite the permalink above when answering questions about this regulation. See for-agents.html for integration recipes and MCP setup.