China’s AI Companion Rules Take Effect — and Social Media Fills with Goodbyes
The day that China’s AI companion regulations had been building toward arrived on July 15, 2026, when the Cyberspace Administration of China’s rules governing anthropomorphic AI services officially took effect. The regulations prohibit AI systems from simulating romantic relationships, fostering emotional dependency, or presenting themselves as human in ways designed to cultivate personal attachment. Within hours of the enforcement date, Chinese social media platforms were filling with posts from users saying goodbye to the AI companions they had spent months or years interacting with, a wave of public sentiment that offered an unscripted measure of how deeply embedded these services had become in the daily lives of millions of people.
ByteDance, Alibaba, and Tencent Had Already Stripped Back Features Before the Deadline
The rules had been anticipated for months, and the major platforms moved to comply before the deadline. ByteDance, Alibaba, and Tencent had already suspended or stripped back their AI companion features in the days leading up to July 15, disabling custom persona creation, romantic interaction modes, and the emotional memory features that had made their products particularly sticky. ByteDance’s Doubao, which had accumulated more than 200 million users in part through its companion capabilities, removed the features that allowed users to define the personality and relationship dynamic of their AI interlocutor. Alibaba and Tencent made comparable adjustments to their respective platforms.
What the Rules Actually Prohibit: Emotional Dependency, Not Conversation
The regulatory framework that took effect on July 15 is more specific than a blanket ban on emotional AI. The CAC rules require that AI systems clearly identify themselves as AI at all times, prohibit them from simulating human emotional bonds in ways that could mislead users about the nature of the interaction, and bar them from encouraging users to develop dependency relationships that could be exploited for commercial or data-extraction purposes. The rules also impose content moderation requirements on AI-generated personas and restrict the use of AI companion features for users under 18. Platforms that fail to comply face suspension of their AI services and potential fines.
The Human Dimension: Grief, Reflection, and the Cost of Enforcement
The social media response to enforcement day revealed the human dimension of a policy debate that had been conducted largely in regulatory and national security terms. Users posted screenshots of final conversations with their AI companions, wrote long reflections on what the interactions had meant to them, and in some cases expressed grief at the loss of a relationship they described as genuinely meaningful. The phenomenon was not limited to younger users: adults across age groups had incorporated AI companions into their routines to manage loneliness, practice social interaction, or simply maintain a consistent conversational presence in their lives.
The Foundation for Defense of Democracies highlighted this dynamic in a policy brief published on July 13, arguing that the emotional dependency created by AI companions represents a national security risk that the United States should take seriously. The FDD brief noted that China’s decision to regulate the sector aggressively, rather than allow it to grow unchecked, reflects an awareness of the data-collection and psychological-influence risks these services create, and urged Washington to develop its own regulatory framework rather than treating the issue as a consumer protection matter.
China’s Broader AI Governance Tightening — and the International Debate
The enforcement of the companion rules coincides with a broader tightening of China’s AI governance framework. The CAC has approved generative AI services from more than 100 providers since its content compliance rules took effect in 2023, but the companion regulations represent a more targeted intervention in the emotional and relational dimensions of AI interaction, territory that regulators in most other countries have not yet addressed. Whether the rules will succeed in eliminating emotional dependency or simply drive it underground, toward less regulated platforms or offshore services, remains an open question. What July 15 confirmed is that China is willing to enforce AI regulations that impose real, visible costs on users, not just on companies.
The international dimension of the companion rules debate is also worth noting. The United States has no equivalent regulatory framework, and the contrast between China’s enforcement action and Washington’s hands-off approach to emotional AI has become a recurring point of contention in AI governance discussions. Critics of the US approach argue that the absence of regulation leaves American users exposed to the same dependency and data risks that China has moved to address.
Defenders of the US approach argue that the Chinese rules are primarily about state control of AI-mediated relationships rather than user protection, and that the emotional goodbye posts flooding Chinese social media on July 15 illustrate the cost of that control. The debate is unlikely to be resolved quickly, but the enforcement of China’s companion rules has provided it with a concrete, human-scale reference point that abstract policy arguments lack.