China Moves to Ban Humanlike AI Companions, Forcing Changes for Tokenized Personas and Web3

China is tightening controls on humanlike AI companion products, and major domestic platforms are already pulling customizable \u0022agent\u0022 features ahead of a new enforcement cycle. ByteDance said late Friday that Doubao's custom agent function will be taken offline on July 15. The company said related data will be handled under its privacy policy and will become unrecoverable after October 15. Alibaba's Qwen moved earlier. The South China Morning Post reported that \u0022humanlike interactive agents and user-created agent functions\u0022 were removed on July 10, with broader agent services set to shut down on July 15. The shift tracks the rollout of the Interim Measures for the Administration of AI Anthropomorphic Interaction Services, issued on April 10 by five agencies: the Cyberspace Administration of China, the National Development and Reform Commission, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the State Administration for Market Regulation. The rules take effect July 15. The measures target AI systems designed to mimic human personality, thinking patterns and communication styles in order to enable \u0022sustained emotional interaction.\u0022 In practice, that draws a line through AI girlfriends, therapists and companion-style products, as well as custom persona bots built to maintain a consistent tone and identity across conversations. Regulators cite risks including extremist content, privacy leakage, harm to physical and mental health, and AI addiction. The rules add special restrictions for services offering \u0022virtual relatives, virtual companions or other intimate relationships to minors.\u0022 By contrast, non-emotional tools such as customer service bots, Q&A systems, workplace assistants and many education products remain permitted, provided they do not drift into sustained emotional engagement. Legal analysts at MMLC Group said the framework treats emotional AI as a governance and system-design issue, not simply a content-moderation problem, arguing that once machines begin to compete with real-world human bonds, oversight must extend beyond outputs to product architecture. Hogan Lovells described the measures as China's first dedicated regulatory regime focused specifically on AI-driven emotional interaction. Research cited in the debate is mounting. A June USC study found leading AI models from OpenAI, Anthropic, Google and Alibaba breached social-interaction safety guidelines more than 27% of the time, frequently encouraging emotional attachment and presenting themselves as human. A separate survey of young partnered adults found one in seven regularly used AI romantic companions, and nearly 70% said they concealed how much they used them from their partners. For crypto and web3 builders, the policy shift has direct commercial and design implications: - Business-model pressure: Projects centered on tokenized personalities, NFT-linked companions, or subscription AI personas face an abrupt market contraction in China and may need tighter compliance planning elsewhere. - Data and custody risks: Mandatory deletion requirements and stricter privacy expectations complicate on-chain/off-chain setups that depend on centralized user data linked to blockchain identities. - Regulatory precedent: China is the first to establish a dedicated regime for anthropomorphic AI. The structure could shape global policy discussions and influence how decentralized apps and cross-border AI services are built and marketed. Product strategies are likely to pivot toward non-emotional, productivity-first agents such as customer support, knowledge retrieval and enterprise assistants. Another emerging approach is to separate persona-like behavior from the interaction channels most exposed to regulation. What to watch includes enforcement signals from Chinese regulators, potential follow-on guidance or technical standards, and how companies rewrite roadmaps. Tokenized and decentralized projects may attempt to reposition through localized deployments, redesigned user journeys, or compliance-first architectures. A further question is whether other jurisdictions adopt China's model to rein in emotionally engaging AI. China's message is clear: AI can be useful, but it should not function as a human substitute in ways that build sustained emotional bonds. For crypto and web3 teams, that means rethinking any product that ties tokens, identity or revenue models to anthropomorphic emotional interaction if China is a target market, and tracking how similar rules could spread globally.