China publishes draft rules on recommendation algorithms

The Cyberspace Administration of China published, for public comment, a set of draft rules governing how internet platforms use recommendation algorithms — among the first attempts anywhere by a government to regulate ranking and personalisation systems directly rather than the content they surface. Comments closed a month later, on 26 September 2021.

The draft required platforms offering algorithmic recommendation services to disclose the “basic principles, purposes and operating mechanisms” of those algorithms to users and regulators, and to give users a convenient way to turn personalisation off entirely — an option platforms had to honour “immediately” on request. It banned algorithms designed to induce addiction or excessive spending, prohibited using algorithms to manipulate account rankings, traffic or search results to create false popularity, and barred pushing content to minors that could encourage unhealthy habits or internet addiction. A separate provision addressed gig-economy platforms specifically, requiring transparency and worker protections in the algorithms that set delivery riders’ routes, pay and penalties — a direct response to reporting that framed such systems as forcing dangerous driving to meet algorithmically generated deadlines.

The draft left its effective date blank, to be filled in once finalised. It was one part of a wider regulatory push that year: Beijing had already moved against Big Tech’s market power and data practices, and would follow with rules on deep synthesis (deepfake) technology and, later, generative AI specifically. The provisions took effect on 1 March 2022 largely as drafted, making China’s algorithm-transparency and opt-out requirements binding well before comparable proposals had progressed through the EU or US legislative process.