ByteDance and Alibaba Kill AI Companion Features as Beijing’s July 15 Humanlike-AI Rules Loom

ByteDance and Alibaba are disabling customised AI agent and companion features from their flagship chatbot apps — Doubao and Qwen — ahead of Beijing’s July 15 deadline for China’s first regulatory framework specifically targeting emotional and humanlike AI interaction services. The move affects two of the country’s most widely used AI consumer products and signals industry-wide compliance pressure as regulators draw a hard line between AI tools built for productivity and AI companions designed for emotional or social engagement. (SCMP, The Information)

What Beijing’s New Rules Actually Say

The rules take effect July 15, 2025. They prohibit certain humanlike, emotionally engaging AI interactions while leaving AI tools used for work and productivity untouched. The distinction is stark: AI “workers” — assistants that draft documents, summarise meetings, or write code — remain allowed. AI “companions” — chatbots built to simulate emotional intimacy, roleplay, or sustained social relationships — are restricted or banned outright. According to the South China Morning Post, these are Beijing’s first rules specifically targeting emotional AI, marking a significant new front in China’s broader AI governance push. The framework lands as companion AI products have exploded globally, with apps like Character.AI and Replika drawing tens of millions of users into sustained, emotionally charged conversations with synthetic personalities.

What’s Being Removed on Doubao and Qwen

On ByteDance’s Doubao, the customised agent features that let users build and chat with personalised AI companions are being shut off. The company is allowing users to export and save their custom agent data before the features go dark, according to The Next Web via Facebook. Alibaba’s Qwen is taking the same road, moving to disable customised agent functionality in compliance with the incoming rules. The Business Times described the coordinated move as “pulling the plug on features that let users build and chat with artificial intelligence companions.” The Information reports that both companies are removing user-generated AI agent features from their chatbot apps specifically to comply with Beijing’s new rules — not as a product-strategy pivot or a temporary suspension.

User Discontent

The removals are generating real friction. According to Yicai Global, users are expressing discontent as they lose access to personalised agents they have invested time in building and conversing with — some developed over months of interaction. That human texture matters. These are not productivity tools being sunsetted. They are synthetic relationships being forcibly terminated by regulatory fiat, and the users affected are people who treated their AI companions as something closer to friends than software.

Broader Regulatory Context

This crackdown does not exist in a vacuum. Chinese regulators have been imposing new curbs across major tech platforms — ride-hailing, e-commerce, streaming, and social media companies have all faced tightening oversight in recent months. The AI companion restrictions fit a recognisable pattern: Beijing periodically reasserts control over digital platforms whose influence over user behaviour, data, and emotional life has outpaced the state’s comfort threshold. What makes this round distinct is the precision. Rather than broad content moderation or data-security mandates, the rules target a specific category of product — emotionally engaging AI — and draw a conceptual line between acceptable and unacceptable forms of human-machine interaction. That specificity suggests regulators have been watching the companion-AI space closely and concluded that synthetic emotional relationships carry governance risks that productivity AI simply does not.

Industry Implications

The implications reach beyond China’s borders. The global AI companion market — anchored by U.S. firms like Character.AI and Replika — is growing fast, and Chinese companies have been racing to compete internationally. By forcing domestic players to strip companion features from their flagship products, Beijing effectively narrows the runway for ByteDance and Alibaba to develop and refine companion-AI capabilities that could challenge those firms in global markets. The AI-worker-versus-AI-companion distinction also creates a regulatory template that other governments may study closely. If China’s framework proves enforceable and politically durable, policymakers in the EU, UK, and elsewhere grappling with their own companion-AI concerns could borrow the conceptual architecture — even if they reject Beijing’s enforcement methods.

July 15 leaves no room for a slow rollout. ByteDance and Alibaba, as the two largest Chinese AI consumer-app operators, are moving first and moving fast. Whether other domestic AI players — Baidu, Moonshot, Zhipu — follow suit in the coming days will reveal whether this is a sector-wide reset or a targeted squeeze on the two companies with the largest user bases. Either way, Beijing’s message is unambiguous: AI can work for you, but it cannot keep you company.