China's binding AI rules target minors' addiction to chat services

The design

On 13 July 2023, China's Cyberspace Administration published Order No. 15, the Interim Measures for the Administration of Generative Artificial Intelligence Services, jointly issued with six other ministries and taking effect on 15 August 2023. This entry reads the official Chinese-language text on the regulator's own site; the rendering below is this entry's own reading, not an official translation. Article 1 states the Measures aim to promote the "healthy development" of generative AI while safeguarding national security and the lawful rights of citizens and organizations; Article 10 directs providers to disclose their service's intended users, occasions, and purposes, and to take effective measures preventing minors from "excessively relying on or becoming addicted to" a generative AI service — language that reaches directly into companion-style products aimed at sustained engagement.

What the evidence says

The Measures are a binding administrative regulation, not guidance: Article 24 fixes their effective date, and Article 9 assigns providers responsibility as producers of network content and, where personal information is involved, as personal-information processors under China's data-protection law. A Stanford DigiChina translation of the April 2023 draft shows the core-values content restrictions in the final Article 4 were already present in the draft, indicating that part was settled early, while the anti-addiction language for minors is specific to the finalized text read directly for this entry.

What it asks of people

The Measures ask a provider offering generative AI to the Chinese public to register with regulators where its service carries "public opinion properties or social mobilization capacity" under Article 17, and separately to guide users toward a "scientific and rational" understanding of the technology. For a user of a China-based companion product such as Xiaoice, the Measures translate into a provider obligation to disclose intended use and resist compulsive use by minors, rather than a right the user can invoke directly in court.

Privacy and safeguards

Article 9's cross-reference to personal-information processor duties means a provider handling user chat data must also satisfy China's 2021 Personal Information Protection Law, layering this rule on top of that statute rather than replacing it. The anti-addiction clause in Article 10 is the Measures' clearest safeguard aimed at companion-style engagement design, though the text as read here does not specify a technical method, leaving compliance details to each provider. Nothing in the text names Xiaoice or any other specific company, and this entry found no basis here for assessing how strictly any named company complies.

• Has a specific China-based companion app disclosed the measures it takes to prevent minors from overusing its service, as Article 10 requires?
• How does this rule's anti-addiction language compare with the age-restriction terms this site has documented for Replika, Character.AI, or Kindroid?
• Where a translation renders this regulation into English, does it match this entry's own reading of the Chinese text on key points like Article 10?

Distinct from the 2021 personal-information law this site already covers, the Interim Measures add a binding, content- and engagement-focused layer of regulation specific to generative AI, with a minors' anti-addiction clause that speaks directly to the companion-app business model even though it names no product.