China's Cyberspace Administration of China (CAC) announced on July 20, 2026 a four-month special campaign — "Clear · Minor Online Protection" (清朗·未成年人网络保护) — ordering platforms to re-engineer recommendation algorithms so they don't "induce addiction or excessive consumption," and to stop using "nearby people" and "people you may be interested in" discovery features that can expose minors to inappropriate contacts. The campaign runs in two phases: a summer-holiday sweep targeting AI-generated cartoon content and payment scams aimed at children, then a second phase standardizing recommendation systems and expanding dedicated minor modes.
The striking detail is what's missing from the announcement: any acknowledgment that this obligation already exists in binding law. Article 18 of the 2022 Provisions on the Administration of Algorithm Recommendation of Internet Information Services — jointly issued by the CAC, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the State Administration for Market Regulation, effective March 1, 2022 — already states that algorithmic recommendation providers "may not use algorithmic recommendation services to lead minors to online addiction" and may not push content likely to harm minors' physical or mental health. That rule has been on the books for four and a half years. The 2026 campaign does not amend it or introduce new penalties; it just orders platforms to actually comply, again.
The steelman: China's minors are online, and the harm case is real
The strongest argument for the campaign isn't abstract. China's minor internet population is enormous and saturated — roughly 193 million minors online as of the last national survey, a 97.2% penetration rate, with short-video consumption among minors climbing from 40.5% to 54.1% between 2018 and 2022. Regulators cite specific, plausible harms: AI tools that let anyone splice violent or grotesque content into children's cartoons "within minutes," per analyst Ding Daoshi; "summer gaming companion" services that lure kids into inappropriate chat; smartwatch apps with tipping and voting features; and — the algorithmic core of this campaign — proximity-based social discovery tools that were never designed with child safety in mind but get used by whoever downloads the app. A platform whose recommender optimizes purely for watch-time has no built-in reason to stop serving a 12-year-old an infinite feed. Some guardrail is a reasonable ask, and Beijing is not alone in reaching for one — the EU's DSA and the UK's Online Safety Act both impose comparable recommender-system obligations on platforms with child users.
But the pattern here isn't new regulation — it's non-enforcement made cyclical
The problem is that "Clear" isn't the first campaign to promise this. In November 2024, the CAC and MIIT launched a three-month drive explicitly targeting "echo chambers" and addictive algorithms, running to a February 14, 2025 deadline, with the same self-examination-and-rectification language regulators are using again now. If that campaign had produced durable compliance, the 2026 version would not need to re-litigate the same algorithmic addiction problem from scratch — let alone stack it on top of the April 2026 rules on AI companion services for minors, which took effect July 15, 2026, just five days before this campaign opened.
This is enforcement by campaign rather than by consequence, and it is a worse regime for platforms and for kids than either strict enforcement or no rule at all. A platform operating under a standing legal duty can build the constraint into its recommender once and budget for ongoing compliance costs. A platform operating under recurring campaigns learns instead that the real cost of non-compliance is a periodic, time-boxed remediation sprint — survivable, predictable, and cheaper than permanent architectural change. That's a rational response to the incentive structure Beijing has built, and it explains why the same violations reappear in each campaign's target list.
What would actually change the incentive
If CAC wants recommenders that don't optimize against children's welfare, the fix is standing, priced enforcement of Article 18 — public fines tied to specific violations, audited algorithm filings (China already requires algorithm registration for public-opinion-affecting services), and liability that survives past the campaign's four-month clock. Campaign-driven enforcement generates headlines and short-term compliance theater; it does not generate the kind of architectural change — actually re-weighting a recommender's objective function — that requires sustained cost pressure. For platforms operating in China and for foreign observers assessing "the China model" of tech governance, the lesson of five consecutive addiction-focused campaigns since 2021 is not that Beijing lacks the legal tools. It's that campaign enforcement is a substitute for the harder, more expensive work of continuous supervision — and substitutes rarely outperform the original.