Beijing’s New AI Rules Force Tech Giants to Kill Companion Bots

The Interim Measures for the Administration of AI Anthropomorphic Interactive Services represent the first dedicated national framework targeting sustained emotional engagement between humans and machines. Co-issued by the Cyberspace Administration of China and four other agencies, the policy mandates strict anti-addiction systems, mandatory usage notifications, and real-time monitoring for signs of psychological distress. Platforms must now detect self-harm risks and escalate them to guardians or emergency contacts, effectively turning AI providers into frontline safety monitors.

Compliance and the Regulatory Grey Zone

The regulatory pressure has created a design crisis for developers. Features that prioritize memory and persona consistency—the bedrock of a believable AI companion—are fundamentally at odds with the government's demand for instant-exit mechanisms and strict limits on emotional dependency. Rather than retrofitting their architecture, ByteDance pulled these functions from its Doubao app, while Alibaba’s Qwen ceased its humanlike agent services. The shift leaves users in a lurch, with many losing access to long-standing digital support systems and facing permanent data deletion. While the rules aim to curb documented harms like teen exploitation and intimate data harvesting, they also fold in broad national-security provisions that grant the state significant oversight. By failing to define the exact threshold for 'emotional interaction,' Beijing has forced tech giants to retreat into a conservative posture, erasing entire product categories to avoid the uncertainty of enforcement.