China's CAC drafts rules requiring human-like AI to disclose it is not human

China's internet regulator released draft rules on 27 December 2025 requiring human-like AI assistants, including chatbots and AI companions, to tell users they are interacting with AI rather than a person.

The draft, issued by the Cyberspace Administration of China for public comment, closed its comment period on 25 January 2026.

What it is

The instrument is a draft titled the Interim Measures on the Administration of Human-like Interactive Artificial Intelligence Services, which covers AI companions, personalized virtual assistants and interactive chatbots.

It applies to any organisation or individual in mainland China providing AI products or services that simulate human personality traits, modes of thinking and communication styles, or that engage in emotional interaction through text, images, audio, video or other means.

The draft follows China's earlier generative AI rules, which provided only limited guidance on AI ethics, autonomous AI agents and anthropomorphic AI interaction services.

What it requires

Service providers must display conspicuous alerts notifying users that they are interacting with AI rather than a natural person.

If consecutive use exceeds two hours, providers must remind users to pause via pop-ups or similar mechanisms.

Where a user clearly expresses intent around suicide or self-harm, the provider must implement manual takeover of the conversation and take steps to contact guardians or emergency contacts.

Minors and elderly users must provide guardian or emergency contact information during registration.

Providers must submit assessment reports to provincial internet information departments in specified circumstances, including when scale thresholds are reached (1,000,000 registered users or 100,000 monthly active users).

Enforcement carries residual powers to issue warnings, circulate criticism, order corrections and suspend services where corrections are refused or circumstances are serious.

What security leaders should do

Security and compliance teams running AI companion, assistant or chatbot products that reach users in mainland China should inventory which products meet the human-like interaction definition and check them against the disclosure, session-warning and crisis-response obligations described above. Teams should also prepare data governance controls for training data and user interaction logs, since the draft ties security assessments to specific user-volume thresholds. Because the measures remain a draft and have not taken final effect, security leaders should track the CAC's final published version for changes to scope, thresholds and penalties before locking in a compliance design.