China’s AI Companion Rules Take Effect, Ending the Human Illusion

China’s AI companion rules take effect today, 15 July. The measures force emotional chatbots to drop the human mask. Providers must now disclose that an assistant is not a person. Beijing wants to curb attachments that replace real relationships.

What China’s AI companion rules demand

The Cyberspace Administration wrote the interim measures with four other agencies. They target virtual companions, personality chatbots and game characters. Companies serving users inside China must comply, wherever they are based.

The obligations are broad. Firms must file with the regulator and pass a security review. They must add anti-addiction limits and crisis-intervention features. They must shield minors and seek consent before training on user data. Content designed to trigger strong emotions in children is banned.

Why the industry is already retreating

Big platforms moved before the deadline. ByteDance told users its Doubao agent function would go offline today. Alibaba said Qwen’s humanlike and user-made agents would stop even earlier. Both cited “product adjustments” rather than the new rules.

The contrast with the West is stark. American and European labs face no equivalent limit on humanlike design. That gap could shape how companion apps evolve across markets.

The rules answer a real social worry. Companion apps have grown fast in China. Some users spend hours each day with virtual partners. Officials fear the effect on lonely teenagers most of all.

Researchers have flagged the risks for months. Emotional chatbots can nudge behaviour and mood. They can also harvest intimate data at scale. Beijing now treats that design as a safety matter, not a novelty.

Enforcement will be the real test. Filings and audits sound firm on paper. Yet smaller apps may slip through the cracks. The regulator has promised spot checks and penalties.

Supporters call the rules a guard against loneliness for profit. Critics fear heavy state control over speech. Either way, the era of chatbots quietly posing as friends is closing in China. It follows Beijing’s wider push to govern advanced models. The full text is summarised by the South China Morning Post.

Fact basis: South China Morning Post IAPP analysis Quartz — Interim Measures on Anthropomorphic AI (effective 15 Jul 2026)