China Wants Platforms to Trace AI-Powered Abuse
China has proposed cyberbullying rules that would require platforms to trace AI-assisted abuse, verify identities and report flagged material to authorities.
Online-safety rules usually focus on removing harmful content. China’s new proposal goes further by asking platforms to identify how AI helped create or spread it.
What happened
China’s cyberspace regulator published a draft cyberbullying law covering abusive content generated or distributed with artificial intelligence.
Platforms would be required to verify users’ identities, trace AI-assisted abuse, remove flagged material and report it to authorities. Potential penalties reach 10 million yuan.
The proposal is open for consultation until 28 August. It is not enacted law, and the final wording may change.
Tracing AI involvement is technically different from ordinary moderation. A platform may need to determine whether a model produced the original content, whether automated accounts amplified it or whether a human edited and reposted it across several services.
Why it matters
The draft would make platforms responsible not only for the visible output but also for the systems that created and distributed it. That could require new provenance tools, identity controls, account-linking systems and records of automated activity.
For AI providers, it may increase pressure to preserve machine-readable signals and cooperate with downstream platforms. Consumer services may also need stricter limits on bots and automated posting.
The rules could reduce some forms of coordinated abuse, but they raise questions about privacy, due process and the scope of government access to identity and platform data.
The bigger picture
AI makes harmful content cheaper to create and easier to personalise at scale. Regulators around the world are therefore moving from reactive takedowns toward accountability for the production and distribution chain.
The hardest problem will be reliable attribution. Content can be copied, edited or generated by open models outside a platform’s control. A requirement to “trace” AI use may be difficult to satisfy consistently without common technical standards.
China’s proposal is an important direction-of-travel signal, but implementation will determine its real effect. Platforms may gain stronger moderation tools while users face more identity verification and surveillance. The policy debate is therefore not only about whether AI-enabled abuse should be controlled, but about how much infrastructure should be built to identify who created it.