The pivot that worked
On May 24, 2026, Taiwan's National Security Bureau (NSB) walked into the Legislative Yuan's Foreign Affairs and National Defense Committee to publicly rebut a wave of Chinese disinformation claiming the United States had "abandoned" Taiwan after the Trump–Xi summit in Beijing. The NSB's tool of choice was not a fine, a takedown order, or a platform liability writ — it was a press conference. That choice reflects a multi-year decision by Taipei to fight disinformation with targeted legal instruments and aggressive counter-speech rather than the European-style comprehensive platform statute it once tried, and failed, to pass.
Almost exactly four years earlier, the National Communications Commission (NCC) approved a Digital Intermediary Service Act (DISA) modeled directly on the EU Digital Services Act, covering any platform above 2.3 million monthly users with algorithmic disclosure, local-representative, and court-ordered takedown duties. Within weeks, the Executive Yuan ordered the NCC back to the drawing board. The agency formally told reporters in September 2022 there was "no timeline" for reintroducing the bill and that it was "back to square one." It has stayed there.
Steelman: why a DSA-style law had a real case
The case for a comprehensive platform statute was not frivolous. Taiwan is, by repeated independent measurement, the most disinformation-saturated democracy in the world per capita, and the Cognitive Warfare Research Centre that the Ministry of Justice's Investigation Bureau opened in January 2024 has documented year-on-year escalation in PRC influence operations. A DSA-style framework would have given regulators predictable levers — risk assessments, algorithmic audits, a single point of contact at every major platform — instead of forcing them to improvise under unrelated statutes. EU regulators argue the same approach has begun to bend platform behavior in measurable ways. Pretending DISA's defenders were closet censors does no one any favors.
What killed DISA — and what replaced it
The bill collapsed for a different reason: the Taiwanese public did not trust the government to wield discretionary content authority. Tens of thousands of citizens voted against DISA on the official Join participation platform, against barely more than a hundred in support. Service providers added free-speech and scope-of-coverage objections. The asymmetry was too large for any democratic regulator to ignore.
Rather than relaunch a smaller version of the same bill, Taipei chose surgical instruments. The Fraud Crime Hazard Prevention Act (FCHPA), enacted in 2024, requires online advertising platforms to label deepfake and AI-generated ads and disclose advertiser identities under Article 31, with fines of NT$200,000 to NT$5 million for non-compliance. Audrey Tang, Taiwan's first digital-affairs minister, has publicly claimed the rule has all but eliminated deepfake political and celebrity ads from Facebook and YouTube inside Taiwan. The Lowy Institute and Taiwan policy researchers have both highlighted the FCHPA as an example of narrow-purpose regulation outperforming broad statutes.
A second pillar is the Anti-Infiltration Act (2020), which criminalises election and policy interference funded by "hostile external forces" — a status definition rather than a content definition, which raises fewer free-speech problems than a generic "disinformation" test would. Both laws sit alongside non-regulatory infrastructure: the Taiwan FactCheck Center, MyGoPen, CoFacts and the LINE-integrated cofact bot, which Taipei has deliberately not funded directly so they retain civil-society credibility.
The Xiaohongshu test case
The boundaries of this approach are not theoretical. In December 2025, Taiwan's Ministry of the Interior invoked Article 42 of the FCHPA — an emergency provision allowing authorities to order ISPs to "suspend analysis service or restrict access" to a site — to impose a one-year DNS-level block on Xiaohongshu (Rednote), citing NT$132.9 million in cumulative fraud losses linked to the platform. Analysis published on I-CONnect noted that while the official rationale was anti-fraud, the underlying concern was cognitive influence on Taiwanese teenagers — exposing a gap between stated and actual purpose.
This is the warning label on Taiwan's playbook. Sector-specific tools work precisely because they are narrow. The moment a fraud statute is stretched to do the work of an unwritten platform-content law, the legitimacy advantage that killed DISA in 2022 begins to erode. A future court challenge to the Xiaohongshu order, likely under Article 11 of the Constitution, will test whether Taipei has overshot.
Lessons for other democracies
Taiwan's experiment cuts against the assumption baked into Brussels, London, and increasingly Washington that the only credible way to govern online speech at scale is one comprehensive statute aimed at "systemic risk." Taipei has produced measurable harm reduction (deepfake ads) and rapid response capability (the May 24 NSB rebuttal) using:
- Statutes scoped to a specific harm — fraud, foreign infiltration — not "disinformation" writ large.
- Real-time government counter-speech instead of compelled platform action.
- Civil-society fact-checkers funded independently of the state.
- Penalties calibrated to commercial actors (ad platforms) rather than ordinary users.
The Xiaohongshu order is a reminder that even calibrated instruments can be over-leveraged when the underlying political pressure exceeds the formal legal authority. But the broader trajectory is the one open-internet advocates should be watching: a frontline democracy under sustained foreign information attack chose not to adopt a sweeping platform liability law, and is fending off disinformation anyway. Legitimacy, not regulatory breadth, is the binding constraint. A smaller law a democracy actually trusts produces more compliance — and more truth — than a larger one it does not.