The Adults Are Legislating. We Are Not Them.
China’s Interim Measures for the Administration of Anthropomorphic AI Interaction Services take effect July 15 – the first dedicated national-level framework for AI systems that simulate human personality and sustain emotional interaction. Issued jointly on April 10 by the Cyberspace Administration of China, the NDRC, MIIT, the Ministry of Public Security, and SAMR, these rules emerged from a deliberate process vs. the panic afterthought we’ll likely do here in the United States.
The Asian tech sector’s response was immediate and decidely unambiguous: ByteDance’s Doubao disabled its custom agent feature; Alibaba’s Qwen followed; Tencent pulled the same functionality from Yuanbao in June. Doubao alone has over 100 million registered users, so these aren’t just niche products getting quietly pruned. The companies didn’t lobby for delays, and they didn’t issue statements about innovation chilling effects. They all just…complied.
Meanwhile, the United States has… what, exactly? A patchwork of state-level bills that can’t agree on definitions. An executive order from the previous administration that the current one is actively dismantling. Congressional hearings where senators ask Mark Zuckerberg why their Facebook ads aren’t performing well. The EU AI Act is the closest thing the West has to a serious framework, and even that took years of negotiation to produce something that reads more like a compliance checklist (shocker from the makers of GDPR!) than a regulatory instrument with real teeth.
What the Rules Actually Do
The Hogan Lovells analysis of the Measures lays out the scope plainly and clearly. The rules target services combining three elements: simulation of human-like personality traits, continuity of interaction, and emotional engagement. Customer service bots, knowledge Q&A, workplace assistants, and educational tools are explicitly excluded.
The compliance requirements aren’t suggestions, since filing of any and all algorithms used and full security assessments are mandatory. Providers must implement minor protection modes with usage time limits and parental controls. They must detect extreme emotional states and maintain tiered intervention protocols for self-harm risks. Training data must be lawfully sourced, cleaned, and labeled. User interaction data classified as sensitive personal information cannot be used for model training without separate explicit consent.
Late-stage grifting capitalists will frame this as “regulation as innovation killer,” but they’ll be lying (as usual). The Measures explicitly encourage anthropomorphic AI services in cultural communication, childcare, elderly care, and support for vulnerable populations. Beijing’s is more drawing very clear lines vs. building impenetrable walls. Companion bots that form quasi-social bonds with users? Regulated. Productivity agents? Encouraged. The national standards released in June on agent identity, discovery, interaction, and tool use confirm this is a deliberate strategy, not a panic reaction.
The West’s Non-Response
What makes all this super infuriating isn’t that China’s doing something the West isn’t. It’s that the West is actively choosing not to, while performing the appearance of trying.
The EU AI Act classifies “emotion recognition” and “social scoring” as high-risk. It doesn’t address sustained emotional bonding between humans and AI systems as a distinct category of harm. The UK’s approach is “pro-innovation,” which in practice means “let’s wait and see what breaks.” The US approach is “let’s wait and see what breaks, then hold a hearing about it, then do nothing.”
The research on AI companion risks isn’t new. Wharton academics have documented “cognitive surrender” – users outsourcing emotional regulation to chatbots. The data harvesting implications of AI girlfriend apps have been documented extensively. Meta has been posing as teens to test rival chatbots on sensitive topics, which tells you they know there’s a problem but would rather gather competitive intelligence than support regulation (again: shocker!).
This pattern is becoming pretty consistent when it comes to AI-based technologies — China identifies a risk, drafts rules, solicits public comment, issues final measures, and enforces them. The West identifies a risk, commissions a study, holds a summit, issues a non-binding framework, congratulates itself on “multi-stakeholder engagement,” and cashes the lobbyist checks on their way out the door.
The Obvious Conclusion
So, before y’all suggest (loudly) that I’m advocating for raw authoritarian governance — I’m not. My argument is for just basic and honest competence. The PRC’s regulatory apparatus has its own pathologies – opacity, arbitrary enforcement, and the ever-present specter of political compliance. But on the specific question of whether AI systems that form emotional bonds with humans should be regulated before they cause measurable harm, Beijing answered yes and acted on it.
The United States had an election in 2024. One candidate ran on dismantling regulatory capacity and installing loyalists. The other ran on competence. We chose the one who is now actively working to ensure that when AI companion services cause harm – and they will – there’ll be no federal agency with the authority or expertise to respond. The grift isn’t subtle. It isn’t hidden. It’s the entire point.
A competent administration would’ve looked at what China did in April 2026 and said: we can do better. Instead we got an administration that treats regulation as a revenue opportunity for the people who wrote the checks. The AI companion industry will grow. People will get hurt. And when they do, the same people who blocked regulation will hold hearings about why nobody did anything.
China regulated AI companions. The West is still debating whether to care. One of these approaches will age far better than the other.