AI Governance Institute
← News
Research2026-05-31

China's Anthropomorphic AI Rules Take Effect July 2026, Setting New Bar for Companion and Interaction Services

What happened

The Cyberspace Administration of China published the Interim Measures for Anthropomorphic AI Services on April 10, 2026, establishing a finalized regulatory framework effective July 15, 2026. The measures apply to services that simulate human-like interaction through AI, including virtual companions, virtual relatives, and emotionally responsive AI personas. Obligations include mandatory disclosure to users that they are interacting with an artificial system, governance over AI-generated content produced within interaction flows, and technical and operational safeguards designed to prevent psychological dependency or harmful emotional influence. Separate and more restrictive provisions apply when such services are offered to minors, effectively prohibiting or severely limiting virtual companion and virtual relative products in that demographic. Operators must also comply with existing Chinese algorithm filing requirements, situating these measures within China's layered AI regulatory stack that already includes the Generative AI Interim Measures, Deep Synthesis Regulations, and Algorithm Recommendation Regulations.

Why it matters

  • ·Regulatory exposure is immediate and layered: organizations operating anthropomorphic AI products in China must comply by July 15, 2026, and non-compliance sits at the intersection of three existing regulatory regimes covering algorithms, deep synthesis, and generative AI, meaning enforcement scrutiny is likely to be coordinated and cumulative.
  • ·The minor-protection provisions create a product-level compliance gate, not merely a disclosure obligation: services that include virtual companion or virtual relative features must either exclude minors categorically or demonstrate compliance with restrictions, requiring age-verification infrastructure and product design controls that many current offerings lack.
  • ·The psychological influence and addiction safeguard requirements are novel in their operational specificity, obligating compliance teams to design and document technical controls against behavioral manipulation that go beyond existing content moderation frameworks and have no direct analog in Western AI regulation yet.

Governance controls affected

SCT-001Anthropomorphic and Companion AI SafeguardsCMP-001Multi-Jurisdiction AI Regulatory Compliance MappingDGC-002PII Handling in AI SystemsSAF-005Harmful Content FilteringMGV-002AI System Intake and Approval Workflow

What to do now

  • ☐Audit all AI products offered in China to identify any service that simulates human-like interaction, emotional responsiveness, or persona-based engagement, and map each against the anthropomorphic AI definition in the CAC measures.
  • ☐Review product eligibility for minor users: determine whether any in-scope service includes virtual companion or virtual relative features and initiate a product decision process on restriction, age-gating, or redesign before the July 15, 2026 effective date.
  • ☐Verify that user-facing disclosure mechanisms clearly and prominently identify the service as artificial at the point of interaction, and document these controls in your AI model registry and product compliance files.
  • ☐Design and implement technical safeguards against addictive design patterns and harmful psychological influence, including interaction time limits, emotional escalation detection, and crisis referral pathways, with documented testing evidence.
  • ☐Confirm that algorithm filing obligations with the CAC are current for all in-scope services, given that these measures sit alongside and cross-reference the existing algorithm recommendation and generative AI regulatory regimes.

What to watch next

Compliance teams should monitor CAC enforcement guidance and any implementing rules issued between now and July 15, 2026, particularly on how the psychological harm and addiction safeguard requirements will be assessed technically. The CAC's broader Draft AI Law, currently moving through the legislative process, may expand or supersede elements of these interim measures, and teams should track whether the National People's Congress advances that legislation before year-end. China's approach to minor protection in AI services is increasingly influencing regulatory discussions in the EU and South Korea, so organizations with global operations should watch for convergence in the design of similar safeguard requirements across jurisdictions.

Stay ahead of stories like this

Get every China AI governance development like this one, plus the rest of the week's developments. Every Thursday.

Powered by Buttondown.

Related Coverage

Enforcement2026-09-06

China Removes 5.6 Million AI-Violative Items in Platform-Scale Enforcement

China's cyberspace authorities removed more than 5.61 million pieces of unlawful or rule-violating AI-generated content and took action against over 49,000 accounts in a nationwide enforcement campaign. More than 2,400 websites and apps were also targeted, with violations covering fabricated false information, AI impersonation, and content harmful to minors. The action demonstrates that Chinese regulators are enforcing AI content rules at operational scale, not just issuing policy guidance.

Corporate Policy2026-09-03

Google's Gemini 3.5 Transcribe Creates Uncharted Compliance Territory for Voice Data

Google announced Gemini 3.5 Transcribe, a speech-to-text model designed for real-time transcription, voice agents, live captioning, and post-call analytics. The model is available for enterprise deployment globally. No safety evaluation documentation or regional use restrictions are publicly surfaced in the available release materials, leaving regulated users to resolve consent, retention, and biometric compliance obligations independently.

Research2026-08-30

Static AI Compliance Documentation Is No Longer Enough, Collibra Warns

Collibra published a practitioner guide on operationalizing AI regulatory compliance across the EU AI Act, US executive orders, and state laws. The guide argues that compliance teams must build a unified AI inventory covering every model, use case, and agent, then encode obligations as automated, evidence-generating controls rather than relying on static documentation. It identifies inventory completeness, policy-as-code, lineage tracking, audit trails, and continuous monitoring as the five pillars of a defensible program.