China's 2023 rules tie AI content labels to an older law

The persona

No single persona anchors this entry. The subject is a Chinese administrative regulation: the Interim Measures for the Management of Generative AI Services, issued jointly by the Cyberspace Administration of China and six other bodies, including the National Development and Reform Commission and the Ministry of Industry and Information Technology. The Measures govern any service that uses generative AI to provide the public inside China with generated text, images, audio, or video, a category broad enough to include the back-end systems behind a synthetic persona's posts.

What the documents establish

The regulation's own text, published on the Cyberspace Administration's site and mirrored on the State Council's official government document portal, records that the Measures were adopted at the agency's twelfth office meeting of 2023 on 23 May, agreed by the other six ministries, and took effect 15 August 2023. Article 4 requires providers and users to uphold core socialist values, avoid content the law lists as prohibited, take steps against discrimination in algorithm design and training data, respect intellectual property and business ethics, and respect others' portrait rights, reputation, and personal information. Article 12 instructs providers to label generated images and video, but it does so by cross-reference: providers must label such content in accordance with China's earlier Provisions on the Administration of Deep Synthesis of Internet Information Services, rather than establishing an independent labelling standard of its own.

Character versus company

The Measures regulate the operator of a generative-AI service, not the character or output it produces. A synthetic persona's posts, images, or video are, under this framework, the responsibility of whichever company runs the generative system behind them; the content rules and labelling cross-reference attach to that provider, not to any fictional voice the persona appears to have. Because Article 12 defers to the separate deep-synthesis rules for the mechanics of labelling, compliance with the Generative AI Measures and compliance with the deep-synthesis labelling regime are two distinct, if linked, obligations.

What to watch

The sources reviewed here establish the Measures' scope, content obligations, and labelling cross-reference as they stood at adoption; they do not show a specific enforcement action against a named virtual-influencer operator. This is an editorial read: because the Measures exempt research and internal enterprise use that does not reach the public, the practical line between an experimental persona and a publicly deployed one may matter more under this regulation than under labelling-only regimes.

• Does a given persona's underlying generative-AI provider operate a public-facing service inside China, bringing it within the Measures' scope?
• Is the persona's generated image or video content labelled in the manner the cross-referenced deep-synthesis provisions specify?
• Has any regulator cited Article 4's content or algorithmic-discrimination obligations against a specific generative-AI operator?

China's Generative AI Measures set content and labelling obligations on the operators of public-facing generative-AI services, with the labelling duty itself pointing back to a separate, earlier deep-synthesis regulation. Confirming whether a specific persona's operator complies with either regime requires checking that operator's own practice against both texts, not just one.