China drafts cyberbullying rules targeting AI-generated abuse with strict platform oversight

China Introduces Comprehensive Draft Rules to Combat Cyberbullying

The Cyberspace Administration of China released draft rules to combat cyberbullying on July 29, marking a significant expansion of Beijing's oversight over online behavior in the world's largest internet community 1

1

2

2

. The proposed regulations specifically target what Chinese regulators call "cyberviolence," a broad category encompassing the online spread of harassing, defamatory, or false information that harms individuals. What distinguishes these measures from previous regulatory efforts is their explicit focus on AI-generated abuse, including deepfakes and synthetic text that can be produced at scale and aimed at individual targets 1

1

.

Platform Obligations and AI-Enabled Abuse Detection

Under the draft cyberbullying rules, China places substantial responsibility on platforms to build detection systems and enforcement mechanisms. Services would be required to develop classifiers that identify cyberviolence, trace and report the risk of cyberbullying content generated or spread using AI technology, and remove and block such content before reporting it to authorities 2

2

. The regulations mandate that platforms offer users a one-click protection feature that blocks incoming messages and comments, addressing the immediate need for victim safety 1

1

. This AI dimension reflects the evolution of online harassment, as older moderation systems were never designed to catch AI-enabled abuse that can generate synthetic harassment at volume.

Real-Identity Verification and Enforcement Machinery

The draft law requires online platforms to verify the real identities of users before allowing them to post information or use instant messaging services, directly addressing the anonymity that has enabled much online abuse in China 2

2

. Platforms must maintain blacklists of repeat offenders, file regular reports on their interventions, and operate early-warning systems to catch pile-ons before they escalate 1

1

. Penalties for abusers range up to account closure, while online service providers that violate the rules could face fines reaching 10 million yuan, approximately $1.5 million, along with potential website or application shutdowns, or revocation of business licenses 2

2

.

Surveillance Concerns and State Control

The same tools designed to detect abuse also identify speakers, and requiring platforms to log, trace, and blacklist users provides authorities with a comprehensive map of online activity, raising concerns about state surveillance 1

1

. Critics of China's internet governance repeatedly point to this tension between protecting victims and surveilling users, with the line between the two drawn by the government. The law would apply not only to cyberbullying activities carried out within China but also to organizations and individuals overseas that target the country, extending Beijing's regulatory reach beyond its borders 2

2

.

Industry Response and Global Context

Some platforms are already adjusting to the anticipated regulations. The question-and-answer site Zhihu has moved to curb anonymous accounts, demonstrating how quickly Chinese companies adapt once a regulatory direction becomes clear 1

1

. Additionally, schools are required to incorporate anti-cyberbullying material into their curricula under the proposed rules 2

2

. These measures fit a broader pattern of Chinese AI rulemaking, as Beijing already requires AI-generated content to be labeled and has regulated human-like AI services, building a framework that treats synthetic media as something to be tracked rather than left unmonitored 1

1

. Other governments from Europe to the United States are circling similar problems around deepfake abuse and synthetic harassment, but China's draft represents one of the more prescriptive regulatory attempts to address AI-generated cruelty 1

1

. The Cyberspace Administration of China is accepting public comment on the draft law until August 28, after which details may shift before the rules take effect 2

2

.