Taiwan's Ministry of Digital Affairs (MODA) has switched on a real-time warning mechanism that flags confirmed fraudulent websites to users the moment they try to visit them — before the site is ever formally taken down. Announced in early August 2026, the system was built with three telecom carriers, Chunghwa Telecom, Taiwan Mobile, and Far EasTone Telecommunications, plus cybersecurity firm Trend Micro and anti-scam app maker Gogolook, whose Whoscall service is one of the region's most widely used call- and link-screening tools (Taipei Times; RTI).
Closing the Window Between Detection and Takedown
The mechanism plugs a specific gap in Taiwan's existing anti-fraud pipeline. When a member of the public reports a suspicious site through MODA's Online Fraud Reporting and Inquiry Portal — live since September 30, 2024 — the report still has to be verified and routed to an ISP before the site actually disappears from the internet. That administrative lag, often hours, is precisely when victims click through and lose money. The new system doesn't shorten the takedown process itself; it inserts a warning screen into that gap. Chunghwa's Hinet broadband, Taiwan Mobile's "Anti-Fraud Battle Police" app, Far EasTone's "Far EasTone Protection Network," Trend Micro's "AI Fraud Prevention Expert" app, and Whoscall now each surface an alert the instant a user attempts to reach a site MODA has already confirmed as fraudulent (Taipei Times).
The scale of the underlying reporting portal explains why this matters. In its first two years, it logged over 1 million fraud reports and led ISPs to take down more than 500,000 confirmed fraudulent sites; a MODA update in July 2026 put the more precise figures at 1,105,170 total reports and 542,082 verified as fraudulent, with average takedown costs down to roughly NT$111 per case (Taipei Times, July 18, 2026). The same reporting also documented steep declines in high-risk scam advertising since a December 2023 baseline — financial-investment scam ads down roughly 99%, impersonation ads down roughly 98% — though those figures track ad-level enforcement under a separate legal regime, not the new ISP alert tool itself.
The Legal Backbone: Liability Without a Content-Removal Mandate
The alert system sits downstream of the Fraud Crime Hazard Prevention Act (詐欺犯罪危害防制條例), which took effect in mid-2024 and requires designated online advertising platforms — Google/YouTube, LINE, Meta, and TikTok — to verify advertiser identity, disclose funding sources, and remove suspected fraudulent ads within 24 hours of notice, with joint civil liability if they don't (official statute text, law.moj.gov.tw). MODA has already tested that liability regime: in May 2025 it fined Meta roughly NT$1 million for failing to disclose advertiser identity on two Facebook ads (Winkler Partners). The ministry paired that statutory stick with direct platform cooperation, running joint anti-fraud sessions with Google, LINE, and Meta as early as August 2024 to build out reporting channels, including a mailbox for public figures whose likenesses are used in scam ads (MODA press release).
The Case For — and the Limit Worth Naming
The strongest case for this kind of intervention is straightforward: financial fraud increasingly moves faster than courts or even administrative agencies can. A victim who clicks a cloned banking or investment site loses money in seconds, and a takedown that arrives a day later helps nobody who already wired funds. Consumer advocates and even free-market technologists broadly accept that some point-of-contact warning is legitimate — the same logic underpins browser "deceptive site ahead" warnings that Google and Microsoft have run for over a decade without much controversy. Taiwan's version simply extends that model from the browser layer to the ISP and app layer, using confirmed reports rather than automated blocklists alone.
What makes this design defensible rather than alarming is what it doesn't do. The alert is opt-in at the product level (Far EasTone customers need a specific subscription; Taiwan Mobile and Trend Micro's versions live inside apps users choose to install) and it warns rather than blocks outright before verification is complete — MODA hasn't handed itself a unilateral DNS-blocking power to silence disputed content. That distinction matters in Taiwan specifically, where the Fraud Crime Hazard Prevention Act's own network-blocking provisions (Article 39's traffic-management clauses) are still sitting in deferred-implementation limbo pending Executive Yuan rules, precisely because blanket blocking authority raises real due-process questions the ministry has so far chosen not to rush.
The honest caveat is that "confirmed fraudulent" is doing a lot of work in a system MODA itself administers, and there's no published independent audit of the verification pipeline's false-positive rate or appeal mechanism for a legitimate site wrongly flagged. A regulator that also runs the alert infrastructure has an obvious incentive to overstate its accuracy in public statistics. Taiwan's approach is still the right template — narrow, opt-in, warning-based, paired with statutory liability for platforms rather than blanket state blocking power — but its credibility depends on MODA publishing verification-error rates with the same enthusiasm it publishes takedown counts.