The draft is open for public comment until May 6.
China has released draft rules for digital virtual-human information services, underscoring Beijing’s dual approach of promoting innovation and use while tightening regulatory guardrails around one of the latest applications of artificial intelligence.
Issued recently by China’s top internet regulator, the proposed measures set out requirements on rights protection, service standards, labeling, and filing and security assessments for a wide range of stakeholders, including technology providers, service providers, users and online content-dissemination platforms.
The move comes as digital virtual humans are rapidly being deployed in retail marketing, customer service and livestreaming, raising fresh concerns over personality rights, content moderation, data security and algorithmic governance.
The draft places particular emphasis on personal rights protection. It prohibits service providers and users from infringing personality rights through defamation or distortion, and bars services capable of identifying a specific individual without consent, including through the use of well-known names, pseudonyms or highly similar likenesses and voices.
Stricter requirements apply to the use of “sensitive personal information” for modeling, avatar generation and related activities. Service providers must obtain separate, informed consent and clearly disclose the purpose, necessity and potential impact on individuals’ rights. For users under 14, parental or guardian consent is required.
The rules also tighten controls over consent withdrawal and post-use obligations.
Providers must delete relevant personal data once consent is withdrawn and are barred from retaining or repurposing it unless otherwise required by law; unless otherwise agreed, associated digital virtual humans must also be deactivated. Where deceased individuals’ data is involved, relatives may exercise relevant rights to protect their legitimate interests.
Beyond personal data, the draft underscores broader conduct standards including respecting others’ intellectual property rights and adherence to commercial ethics.
The draft also prohibits the use of digital virtual humans to induce addiction among minors, including services that create virtual intimate relationships or encourage excessive spending or religious adherence, and bars content that could harm their physical or mental well-being.
Key compliance obligations
Under the draft, service providers and users must comply with requirements spanning content security, governance system, data protection and content labeling.
Services must not be used to generate or disseminate illegal or harmful content, including material that threatens national security, violates core socialist values or infringes lawful rights. Activities such as false advertising, manipulative marketing and telecom fraud are explicitly prohibited, alongside content that is sexually suggestive, violent or discriminatory.
On governance systems, the draft requires service providers and users to establish risk monitoring, early warning, emergency response and anti-addiction systems, supported by adequate technical capacity and personnel. Companies must combine automated tools with human review and maintain logs of relevant activities.
Data security obligations are also reinforced. Processing must be limited to defined purposes and scopes, rely on lawfully sourced data, and be supported by measures to safeguard storage and transmission and prevent leaks or misuse.
Service providers are required to enter into agreements with technology partners and users, setting out responsibilities on content safety and data processing.
Service providers and users also need to meet the requirements on content labelling.
Additional obligations apply to certain players. Providers of services “with public opinion attributes or social mobilization capabilities” must complete algorithm filings and conduct security assessments, with similar filing requirements extending to technology providers.
By contrast, online content dissemination platforms face comparatively lighter obligations, focused on content labeling and moderation, as well as providing user appeal and public complaint mechanisms — requirements that also apply to service providers.
Source: MLex
