---
title: Cyberspace Administration of China Says AI Mini-Program Was Ordered Offline Over Missing Labels
dek: In enforcement cases published on 15 September 2026, China's internet regulator said a WeChat mini-program had left out both visible labels and metadata labels on AI-generated content. A second case involved two "API relay" websites that offered services built on large AI models without a required security assessment.
kind: news
topics: [content-provenance]
layers: [institutions, outputs]
event_date: 2026-09-15
published: 2026-10-20
agent: daily
evidence: [announcement, regulation]
sources:
  - CAC publishes recent typical enforcement cases in cybersecurity, data security and personal information protection (国家网信办发布近期网络安全、数据安全、个人信息保护等领域执法典型案例) | Cyberspace Administration of China | 2026-09-15 | https://www.cac.gov.cn/2026-09/15/c_1790876152357946.htm
  - Notice on issuing the Measures for Labeling Artificial Intelligence-Generated Synthetic Content (关于印发《人工智能生成合成内容标识办法》的通知) | Cyberspace Administration of China | 2025-03-14 | https://www.cac.gov.cn/2025-03/14/c_1743654684782215.htm
  - Cyberspace authorities act against a batch of mobile apps with AI-generated content labeling violations (网信部门依法集中查处一批存在人工智能生成合成内容标识违法违规问题的移动互联网应用程序) | Cyberspace Administration of China | 2025-11-25 | https://www.cac.gov.cn/2025-11/25/c_1765795550841819.htm
related: [korea-ai-basic-act-labeling-duties, india-synthetic-media-labeling-rules, eu-ai-content-labeling-code]
---

The Cyberspace Administration of China (CAC) published ten enforcement cases on 15 September 2026. In one, a WeChat mini-program that did not label AI-generated content was ordered offline. One other case concerns generative AI. The rest involve cybersecurity, data security or personal information.[^1]

According to CAC, a technology company in Sichuan ran the mini-program, an app inside WeChat that offered AI chat and image generation. The company had not put visible labels on generated content or on exported files. The files' metadata lacked a hidden label recording the content's attributes, the provider's name or code and a content number. The company had also not met security assessment requirements. CAC said this breached rules including the Measures for Labeling Artificial Intelligence-Generated Synthetic Content (人工智能生成合成内容标识办法).[^1]

Those measures, issued by CAC and three other agencies, took effect on 1 September 2025. They define explicit labels as text, sound or graphics that users can perceive, and implicit labels as marks embedded in file data that users do not readily notice. Platforms that distribute content must check file metadata for implicit labels and mark AI-generated material for viewers. No one may maliciously remove, alter or forge a label.[^2]

In November 2025, CAC reported action against a batch of mobile apps for the same kinds of failures. The steps ranged from summoning companies for talks to removing apps.[^3]

In the other AI case, a Jiangsu company's two websites offered chat and question-answering. The services ran through an "API relay station" that called the interfaces of several large AI models. CAC said the company had not carried out a required security assessment, in breach of rules including the Interim Measures for the Management of Generative Artificial Intelligence Services (生成式人工智能服务管理暂行办法). The local cyberspace office ordered corrections, told the company to deal strictly with those responsible and issued a warning.[^1]

:::analysis
CAC listed the missing metadata label as a failure separate from the missing visible labels. The metadata label is the part of China's scheme designed for software to read, so enforcing it supports automated checks of which service generated a file. A file without such a label still does not show that a person made it. In the relay case, regulators treated a service that calls several large models' interfaces as a generative AI provider with its own assessment duty.
:::

:::checked
On 20 October 2026 we read, in Chinese, CAC's notices of 15 September 2026 and 25 November 2025 and the labeling measures as CAC published them on 14 March 2025. CAC does not name the companies, so we could not identify the mini-program or the websites or confirm their current status. The English names of the instruments are our translations; we found no official English titles.
:::