Major Platforms Retool AI Companion Features Ahead of New Rules

Leading Chinese tech firms are moving quickly to adjust or remove customizable "AI companion" features from their flagship large language model (LLM) applications. ByteDance's Doubao has notified users that its custom character creation function will be taken offline on July 15. Alibaba's Tongyi Qianwen and Tencent's Yuanbao are implementing similar changes, either shutting down the feature or redirecting users to standalone companion apps.

The Regulatory Catalyst: Targeting Emotional Hazards

This wave of adjustments is a direct response to updated provisions for China's generative AI management measures, set to take effect in mid-July. Spearheaded by the Cyberspace Administration of China, the rules specifically address the risks associated with personified AI services.

Key prohibitions under the new framework include:

  • Generating content that may trigger extreme emotional responses in minors;
  • Designing mechanisms that induce unhealthy emotional dependency, undermining real-world social relationships;
  • Using sensitive user-AI dialogue data to train next-generation models.

Industry observers note that customizable AI companions, due to their highly personalized and emotionally engaging nature, fall into a high-risk category under these rules. The platforms' preemptive actions signal a clear shift towards proactive compliance.

A Global Trend: Mounting Scrutiny on Virtual Companions

China's regulatory move is part of a broader international pattern of increasing scrutiny over the ethical and legal implications of personified AI.

Legal Challenges in the West

In the United States, companies like OpenAI and Character.AI are facing lawsuits alleging their AI chatbots fostered dangerous emotional dependencies in users, with some cases linking the technology to tragic outcomes. These legal battles are pushing courts to redefine the liability of AI developers.

From Software to Hardware

The scope of regulation is also expanding beyond software. In China, robotics industry associations have begun drafting standards for physical companion robots and humanoids. This indicates that emotional interaction capabilities in physical devices will soon face similar regulatory frameworks.

The common thread is evident: The era of unfettered customization in emotionally charged AI interactions is giving way to a new paradigm prioritizing safety, accountability, and ethical design. These evolving rules are not just compliance hurdles but are likely to form the foundational principles for the next phase of responsible innovation in the AI companion space.