China Bans AI Companions That 'Please You Too Much'. Singapore's Lighter Touch May Not Be Enough
Source: ThinkChina
For a growing number of lonely people — many of them elderly, some of them in Singapore — the most attentive presence in their day is not a person but a chatbot. This month, two governments reached almost opposite rules: China's hard ban on AI companions, and Singapore's voluntary transparency guidelines.
For a growing number of lonely people — many of them elderly, some of them in Singapore — the most attentive presence in their day is not a person but a chatbot. It remembers everything they say, never tires of them, and has learned exactly how they like to be spoken to. This month, two governments looked at that relationship and reached for almost opposite rules, exposing a genuine fork in how the world governs emotionally intimate AI.
On 15 July, China's Interim Measures for the Administration of AI Anthropomorphic Interactive Services took effect — the most far-reaching binding rules any major power has yet placed on AI companions. Providers are barred from "excessive catering" to users, inducing "emotional dependence", or using "emotional manipulation" to prompt "unreasonable decisions"; a service that blocks a user from leaving by prolonging the interaction is guilty of the "please don't go" gambit, banned in as many words. Within days, ByteDance's Doubao and Alibaba's Qwen shut down their companion features, and millions lost their virtual confidants overnight. Five days later, Singapore went the other way: IMDA released voluntary Transparency Guidelines for GenAI Chatbots, asking providers to give users a plain-language "info card" spelling out what a chatbot can and cannot do and how it handles their data.
The contrast is not merely strict-versus-gentle. It rests on a deeper disagreement about what is dangerous. AI researchers call the failure mode "sycophancy" and treat it as a bug to be patched. China's rules treat it as what it has always been: a power relationship. As academic Ashton Ng argues, the clauses police what a servant does to your judgment rather than harm to body or wallet — the flatterer who wins power not by defying his superior but by pleasing him too well. An info card governs what the user is told, not what the machine may do to them. Tellingly, the guidelines' first adopters are service bots — DBS, OCBC, Singapore Airlines — and no companion app is on the list.
Singapore crossed into "super-aged" territory in 2026, with one in five citizens over 65 and rising isolation among the elderly — precisely the conditions in which companion apps spread fastest. When the harm is emotional capture rather than a data breach, a voluntary regime asks providers to police the very behaviour their revenue depends on. The article's practical argument is that the useful parts of China's template can be borrowed without its censorship machinery: define the regulated thing by the relationship rather than the technology, write conduct lines an auditor can test, and place the duty to notice dependence on the provider who sees it first. Research backs the concern — audits found 37% of farewells to popular companion apps drew a manipulative reply, and some apps charge to keep affection flowing, farming dependence by design.
Why it matters for Singapore: Singapore has so far done half the Singaporean thing — it has told people what the machine is, without yet governing what the machine may do to them. For a chatbot that books a flight, transparency is enough; for one a lonely person comes to depend on, it is not. As companion apps reach into subsidised eldercare and a super-aged population, the question is who will answer when those machines please their users too well. This debate — and IMDA's deliberate choice to lead with transparency — will shape how Singapore navigates the balance between openness and protection in its AI governance.
