--- title: "Why Chinese apps now stamp \"AI生成\" on everything you see" date: 2026-09-28 category: Policy & Governance site: NeuroAI canonical: https://neuroai.site/a/na-policy-ai-content-labeling language: en --- # Why Chinese apps now stamp "AI生成" on everything you see > China's Measures for Labeling of AI-Generated/Synthetic Content (《人工智能生成合成内容标识办法》) took effect on 1 September 2025, requiring explicit and hidden labels on AI-made text, images, audio, video and virtual scenes. The rule binds service providers, platforms and users in one framework. Scroll a Chinese short-video feed now and you will notice small tags — "AI生成" (AI-generated), "疑似AI制作" (likely AI-made). They are not a platform's polite suggestion. A binding national rule put them there, and the platforms had to comply or risk being taken down. For anyone publishing in China, the change is practical, not philosophical: synthetic content now carries a digital ID whether the creator likes it or not. ## The rule that switched on The **Measures for Labeling of AI-Generated/Synthetic Content (《人工智能生成合成内容标识办法》)** were issued jointly by the **Cyberspace Administration of China (国家网信办, CAC)**, the **Ministry of Industry and Information Technology (工业和信息化部, MIIT)**, the **Ministry of Public Security (公安部)** and the **National Radio and Television Administration (国家广播电视总局, NRTA)**. Dated 7 March 2025 and released mid-month, they took effect on **1 September 2025**. Running alongside them is a mandatory national standard, the **Information Security Technology — Labeling Method for AI-Generated/Synthetic Content (《网络安全技术 人工智能生成合成内容标识方法》)**, effective the same day. Together they turn "please label" into "you must." ## Two kinds of label The rule splits labeling into two: - **Explicit labels (显式标识)** — ones a user can see: a text note such as "AI生成" at the start or end of an article, an icon on an image, a voice or on-screen prompt on audio and video. - **Hidden labels (隐式标识)** — metadata written into the file itself, carrying an **"AIGC" marker** plus provider and trace information, so content can be traced even after it is downloaded or re-shared. The point is coverage: a casual viewer sees the explicit tag, while platforms and regulators can read the hidden one. ## Who has to do what The measures assign duties across the chain: - **Providers** must add explicit labels to generated outputs and embed hidden labels in file metadata; exports and downloads must keep the label. - **App stores and platforms** must check at listing whether a service generates synthetic content and verify its labeling. - **Users** who post AI content must declare it and use the provider's labeling function. - **No one** may maliciously delete, alter, forge or hide these labels. By 1 September, major platforms — **Tencent (腾讯)**, **Douyin (抖音)**, **Kuaishou (快手)** and **Bilibili (哔哩哔哩)** — had already published detailed rules and turned on metadata read/write. ## Why it exists The measures sit on top of earlier rules: the **Interim Measures for Generative AI Services (《生成式人工智能服务管理暂行办法》)**, effective 15 August 2023, and the **Provisions on Deep Synthesis (《互联网信息服务深度合成管理规定》)**, effective 10 January 2023. Those set the ground; the 2025 measures add the "how" of labeling. The stated aim is to blunt fake news, identity impersonation and the slow erosion of trust in what people see online. ## What it cannot fix Labeling is a flag, not a verdict. It depends on providers complying, and hidden metadata can be stripped by third-party tools; explicit tags can be ignored by viewers. Enforcement is strongest inside China's jurisdiction and weaker across borders, where a relabelled or re-exported file may lose its trace. And a label tells you *how* content was made, not whether it is true. ## What readers can do now - **If you publish in China**, label AI content and keep the metadata on export — major platforms now reject or downrank unlabeled synthetic posts. - **If you consume content**, look for the explicit tag (显式标识); treat sensational clips with no label and no source with extra caution. - **If you build products**, bake labeling into your export pipeline now. It is mandatory, not a nice-to-have. ## Honest limitations Core facts (the Measures' four issuing bodies — CAC, MIIT, Ministry of Public Security, NRTA; document date 7 March 2025; effective 1 September 2025; the companion mandatory national standard effective the same day; the explicit/implicit labeling framework; duties of providers, platforms and users; the 2023 Interim Measures and Deep Synthesis Provisions as prior bases; platform compliance by Tencent, Douyin, Kuaishou, Bilibili) come from gov.cn, cac.gov.cn, miit.gov.cn and Xinhua/news.cn. The effective date and issuing bodies are confirmed by both the official text and Xinhua. The rule applies within China's jurisdiction; cross-border enforcement is limited. No RMB amounts are stated. Analysis is current to 28 September 2026. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-policy-ai-content-labeling Free to quote with attribution and a link to the original.