What happened
China's Cyberspace Administration (CAC), jointly with NDRC, MIIT, MPS, and SAMR, published draft 'Interim Measures for the Administration of AI Anthropomorphic Interaction Services' on April 3, with public comment open until May 6, 2026. The measures take effect July 15, 2026 and impose security assessments, mandatory AI identity labelling, emotional-state monitoring, and dependency-intervention obligations on providers of AI companions, chatbots, and emotionally interactive digital services.
Why it matters
Compliance is mandatory for any service with 1M+ registered users or 100K+ monthly active users in China; operators must file security assessment reports with provincial-level CAC before going live or upon reaching those thresholds. The rules also embed psychological safety obligations, raising the bar significantly beyond traditional cybersecurity compliance.
Action needed
Organisations offering interactive or emotionally engaging AI services in China should assess whether user thresholds trigger mandatory security assessment filing obligations before the July 15 deadline; review the draft now and consider submitting comments before May 6.