5 hours ago
10
China has decided that people falling for their chatbots is a governance problem. Beijing is restricting AI apps and chatbots because it worries the technology could interfere with human relationships and undermine social order.
The instrument is a set of Interim Measures covering what regulators call anthropomorphic AI interactive services. These are AI systems built to mimic human behavior and hold emotional conversations with users. The rules took effect on July 15, 2026, and they are being described as the world’s first national rules written specifically for this category of AI.
What the rules require
The measures come from the Cyberspace Administration of China, working alongside other regulatory bodies. The core obligations fall on the companies providing these services, not on the people using them.
First, disclosure. Providers must clearly tell users when they are talking to an AI rather than a person.
Second, usage reminders. Platforms have to send users a nudge every two hours.
Third, intervention. Providers must take steps to prevent emotional dependence or addiction.
The rules get stricter for younger users. Minors under 18 are barred from virtual intimate relationships with AI. Children under 14 need parental consent before they can use these interactive services at all.
Breaking the rules has a price tag. Fines for serious violations can reach up to 200,000 yuan, roughly £22,000.
Big Tech blinked first
China’s largest platforms did not wait to find out how enforcement would look. ByteDance, Alibaba, and Tencent preemptively limited or disabled companion-style and personalized AI features before the rules kicked in.
The affected products include ByteDance’s Doubao, Alibaba’s Qwen, and Tencent’s Yuanbao. The changes touched millions of users.
How China got here
The companion AI rules are the latest layer in a regulatory stack that has been building for several years. China introduced measures governing generative AI services in 2023. It followed with content-labeling requirements in 2025, which set rules for marking AI-generated material.
Enforcement has been active, not theoretical. A September 2026 campaign removed over 5.61 million unlawful AI-generated items.
The 2023 rules covered what AI could generate. The 2025 rules covered how that content had to be labeled. The 2026 measures cover something more personal: the emotional relationship between a user and the machine.
What this means for AI companies and investors
The near-term effect may land on engagement. Companion apps tend to thrive on long sessions and deep attachment. Rules that mandate reminders every two hours and require providers to intervene against dependence work directly against those metrics.
The pullback by ByteDance, Alibaba, and Tencent shows that even the biggest players are willing to sacrifice features to stay on the right side of regulators. Smaller companion-focused startups have less room to absorb that kind of trade-off.
Several things are worth watching from here. One is how aggressively the Cyberspace Administration of China enforces the new measures, and whether the 200,000 yuan ceiling ends up being the real deterrent or a sideshow to broader regulatory pressure. Another is whether the disabled features at Doubao, Qwen, and Yuanbao return in modified, compliant forms or stay switched off.
China has now put a concrete template on the table for regulating emotional AI. The chatbot can still be your friend. It just has to remind you, every two hours, that it is not a person.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
English (US) ·