China's internet regulator ordered a mini-program offline because it did not label AI-generated content, one of 10 enforcement cases it published on 15 September [1].
The app, run inside WeChat by a company in Sichuan, offered AI chat and image generation. It added no visible labels, no hidden labels in file metadata, and no labels on exported content, the Cyberspace Administration of China said [1].
A company in Jiangsu was warned for running two "API relay" websites that passed users' requests to several large AI models without the required security assessment [1].
What it covers
The other eight cases are about security and personal data. Among them [1]:
- a hospital in Henan whose firewalls and other security tools had stopped working when their licences expired, leaving 700 vulnerabilities, 73 of them high-risk [1].
- a Chongqing property company that collected over 5,000 customers' face images without separate consent [1].
- a Shanghai company that sent users' names, phone numbers, emails and passwords abroad without the required security assessment [1].
- a company whose data-management system had over 2,400 internal documents stolen, through an open-source search engine with an unauthorised-access flaw [1].
- a former employee who planted malicious code in an app with about 350,000 users [1].
What it does not tell you
How large the fines were, or who the companies are. The regulator describes the cases without naming them or giving amounts [1].
How often the AI labelling rules are enforced. These are cases the regulator chose to publish as examples [1].
Comments