China's draft cyberbullying bill adds AI to the target list, with reported $1.5M fines and overseas reach

China's internet regulator released a draft "Law on Countering Cyberviolence" on July 29 that reads like a cyberbullying bill, but its operational move is layering a behavior-and-platform-duty statute on top of the country's existing AI governance stack. The Cyberspace Administration of China (CAC) opened the draft for public consultation through August 28, and the most concrete new ingredient is explicit capture of AI-enabled abuse: generators and spreaders of AI-made or AI-amplified harassment fall inside the same platform-duty regime as human offenders.

The framework lands on platforms first. Operators must verify users' real identities before they can post or message, monitor and remove cyberbullying content, and report serious cases to regulators. A specific carve-out treats AI-created or AI-spread material as a flagging trigger: when a platform's systems identify bot-generated or algorithmically amplified harassment, the duty to act is the same as for any other cyberviolence report, according to the China Law Translate working translation by Jeremy Daum at Yale Law School's Paul Tsai China Center.

Penalties reported in coverage of the draft reach up to 10 million yuan (about US$1.5 million) per violation, with the harsher end of the scale reserved for platforms that fail the real-name verification, monitoring, or AI-content duties. The same package lets the regulator order site or app shutdown and, in persistent cases, revocation of a company's internet business license, according to the ANI wire pickup of the consultation text. The 10 million yuan ceiling has not yet been independently confirmed against the CAC's penalty chapter and should be read as the figure in circulation, not a settled statutory cap, until the final text is published.

Article 3 asserts that the law applies to cyberviolence carried out within the People's Republic of China and to overseas organizations and individuals that target Chinese users or platforms, per the China Law Translate working translation. That is a familiar pattern in Chinese internet legislation, but pairing it with the AI-abuse clause is new: an overseas user or operator whose model or bot network drives harassment into China now sits inside the same compliance perimeter as a domestic platform.

The freshest human-scale detail is buried in the school provisions. The draft requires primary, middle, and high schools to add cyberbullying prevention to their teaching programs, according to Guancha's read of the consultation text. It is a small clause next to the penalty chapter, and it does the most to reframe the law as a behavior statute aimed at the whole online stack, not just a regulator-versus-platform enforcement tool.

The architecture matters more than the label. China already has generative-AI measures, deep-synthesis rules, and interim algorithm regulations that govern how models are trained, labeled, and disclosed. This draft does not replace any of them. It sits on top and governs what people and platforms do with AI on the content side. If a model is misused to manufacture insults, harass a user, or seed a rumor campaign, the case is no longer handled only under the existing AI rules; it is also a cyberviolence case with platform monitoring duties, real-name traceability, and the 10 million yuan (about US$1.5 million) ceiling behind it.

Civil society groups have criticized the real-name verification requirement as a privacy and speech cost that goes beyond anti-harassment. The extraterritorial clause will draw attention from foreign platforms that do not currently see themselves as inside China's compliance orbit. The school curriculum clause will draw a different critique: that prevention education is being asked to do enforcement-adjacent work that platforms and parents are better placed to handle. None of those critiques are settled, and the August 28 comment window is the formal channel for them.

The falsifier is narrow. If the final law drops or waters down the real-name verification, the AI-monitoring carve-out, or the extraterritorial clause, the overlay framing weakens and the statute reads as a narrower cyberbullying bill. Until then, the operational read is straightforward: a behavior-and-platform-duty overlay using cyberviolence as a doorway, with the AI-abuse hook as the freshest entry point and 10 million yuan (about US$1.5 million) as the named price of getting it wrong.