ChinaInterim Measures for the Management of Generative AI Services, Article 12
Interim Measures for the Management of Generative AI Services (生成式人工智能服务管理暂行办法), 2023, Art. 12
A citation is an address, not a summary. The first part names the law; what follows narrows it to the exact section, article or paragraph.
What follows is LexLint's own research summary of this law, not legal advice.
In force since 15 August 2023.
An AI transparency rule binding public and private bodies.
As of 14 August 2026.
What it requires
- Label generated images, video, and comparable output if you offer generative AI services to the public in China
- Follow the Deep Synthesis Provisions' Articles 16 and 17 for how to label, rather than a separate standard of your own
What it reaches
How the hook was established
express
What makes it apply
Market targeting
What it makes you log
Logging duty
Article 12 requires a generative AI provider to label generated images, video, and comparable output in accordance with the Deep Synthesis Provisions' Articles 16 and 17; it states no logging or record-keeping duty of its own. Article 14 of the same Measures separately requires a provider to preserve records on discovering a user's illegal activity, and Article 19 gives the competent authorities an on-request inspection power over training-data and algorithm information; neither sits within Article 12's own text.
- Kind
- None
- As of
- 22 September 2026
Who checks it
Audit expectation
none
What this law does
Drafted with AI from the cited sources under the direction of UnGovr staff. UnGovr holds editorial responsibility for this page.
Providers of generative AI services to the Chinese public must label generated content such as images and video in accordance with the Deep Synthesis Provisions' Article 16 and 17 labeling rules. This is a cross-reference duty binding generative-AI providers specifically to the general deep-synthesis labeling regime, not a freestanding labeling standard of its own.
Article 7 of the same instrument governs training-data lawfulness and has no labeling content; it is training-data territory outside this topic's scope, correcting an earlier derivation that had cited it here.
When LexLint raises it
generates_contentdeploys_chatbot
Read the law
Every line above is drawn from the primary source linked here, read on the date shown. This is a research summary, not legal advice.