China drafts strict platform rules to tackle AI-generated online abuse
Beijing’s proposed regulations would force technology companies to build extensive surveillance and moderation systems to detect synthetic harassment, setting a highly prescriptive global benchmark for platform liability.
The Cyberspace Administration of China released draft regulations on 29 July requiring internet platforms to detect and block AI-generated harassment. The proposed rules target synthetic text and deepfake images used in coordinated online abuse, placing the primary compliance burden directly on technology companies.
Under the draft, services must deploy classifiers to identify abusive content and provide users with a one-click protection tool to block incoming messages. Platforms are also required to maintain blacklists of repeat offenders, operate early-warning systems to prevent coordinated pile-ons, and submit regular intervention reports to regulators.
Technology firms are already adjusting their operations in anticipation of the final text. The question-and-answer platform Zhihu has begun restricting anonymous accounts, reflecting how quickly Chinese companies adapt when regulatory priorities become clear.
For global investors and technology markets, the draft signals a massive expansion of compliance costs for companies operating in the region. Firms will need to invest heavily in new moderation infrastructure to meet the expectation that they monitor all user-generated content and answer for any synthetic abuse that slips through.
The regulations also highlight a fundamental tension in Beijing’s approach to internet governance. While framed as a necessary shield against cyberbullying that has driven victims to severe harm, the rules effectively eliminate online anonymity by forcing platforms to log, trace, and identify the individuals behind pseudonymous accounts.
This prescriptive approach contrasts with the ongoing regulatory efforts in Europe and the United States, where lawmakers are also struggling to address AI-enabled cruelty. However, China’s framework goes further by explicitly integrating these moderation tools into a broader state surveillance apparatus, treating synthetic media primarily as a tracked commodity rather than just a consumer protection issue.
The draft remains open for public comment, meaning specific technical requirements for AI detection could shift before finalisation. Regardless of the final wording, the regulatory trajectory is firmly established, demanding that platforms build the infrastructure to police AI-generated abuse at an unprecedented scale.
