The Numbers Behind the Block
On December 4, 2025, Taiwan's Ministry of the Interior ordered internet service providers to stop resolving the domain for Xiaohongshu (RedNote), the Chinese lifestyle-and-shopping app used by an estimated 3 million Taiwanese users, citing 1,706 fraud cases and losses exceeding NT$247 million (~US$7.9 million) tied to the platform, plus a failing grade on all 15 items of a national cybersecurity review (Taipei Times). Nine months in, Interior Minister Liu Shih-fang reported that fraud cases linked to the platform have fallen roughly 80% and financial losses roughly 55% since the block took effect — and signaled Taiwan will likely extend the one-year order if Xiaohongshu's Shanghai-based operator still won't designate a Taiwan legal representative or submit to security review (RTI).
The Case for the Block
The strongest version of the government's position is not about content moderation — it's about jurisdiction. Article 42 of the Fraud Crime Hazard Prevention Act lets competent authorities order ISPs to suspend DNS resolution during a designated fraud-prevention emergency (Laws & Regulations Database), and Taiwan's Ministry of Digital Affairs runs a standing DNS RPZ procedure specifically for illegal-website takedowns, with administrative-appeal and litigation remedies built in for affected parties (Winkler Partners). Xiaohongshu wasn't blocked for what users posted; it was blocked because its operator, unlike Meta, Alphabet, ByteDance, or Line, ignored 52 days of outreach and never appointed the local legal representative every other major platform maintains under Taiwan law. That is a narrower, more defensible trigger than most DNS-blocking regimes worldwide, and a platform with zero legal presence genuinely is harder to compel through ordinary enforcement — subpoenas, fines, and takedown notices all require someone in-jurisdiction to serve.
Where the Evidence Gets Thin
The 80%/55% figures are the headline, but they aren't a controlled comparison — they're a before/after on one platform with no counterfactual. Liu's own numbers undercut the causal story: nationwide fraud cases fell about 20% and losses about 46% over roughly the same August 2024–August 2026 window, meaning Taiwan's broader anti-fraud push (which includes new financial-institution and telecom obligations under the same Act) was already driving a substantial decline before and independent of the Xiaohongshu block (RTI). A platform-specific drop that outpaces the national trend is suggestive, not dispositive — it could equally reflect users migrating fraud activity to a platform that's no longer being counted as "Xiaohongshu-linked" once the DNS block forces them onto VPNs, alternate DNS resolvers, or other apps entirely. Taiwan hasn't published a methodology showing how fraud is attributed to a specific platform, or whether investigators can still trace it once users route around the block — which most technically capable users can do in minutes.
The Precedent Problem
Opposition lawmakers have already warned the measure risks eroding "internet and speech freedoms in Taiwan" even while accepting the fraud rationale (Focus Taiwan).
That critique deserves more weight than the government has given it. Article 42 has no duration cap, no compliance benchmark that triggers automatic restoration, and no requirement that the emergency designation be reviewed by a court before it takes effect — only after, via appeal. A platform millions of people use for shopping and lifestyle content, not primarily fraud, can be DNS-blocked indefinitely on a rolling one-year basis with the sole off-ramp being voluntary compliance from a foreign operator that has already shown no interest in complying. That's a structurally different risk than a takedown of a single fraudulent listing or scam site — it's infrastructure-level blocking of a mass-communication platform, and the standard for extending it should rise, not stay flat, the longer the block persists without the operator engaging.
What Proportionate Would Look Like
Taiwan doesn't need to abandon Article 42 to fix this. Before extending the order past its one-year mark, the Ministry of the Interior should publish the attribution methodology behind the 80%/55% figures, net out the effect of the concurrent national anti-fraud measures, and set a public, time-bound compliance standard — legal representative appointed, security review passed — that would trigger automatic unblocking rather than leaving restoration entirely to ministerial discretion. Fast administrative action against an operator that ignored 52 days of outreach is a reasonable emergency response. Renewing an open-ended block on a 3-million-user platform based on a statistic that the government's own data partially contradicts is a different, much less defensible thing — and the gap between those two is exactly where evidence-based regulation is supposed to do its work.