Taiwan Taiwan anti-disinformation platform law

Taiwan Is Policing Election Disinformation With Criminal Law Because Its Platform Rules Cannot Reach Foreign Platforms

Police have opened 184 disinformation cases before the 28 November vote, while the deepfake takedown mechanism has gone unused. The gap is jurisdiction over platforms.

Taiwan's Election Disinformation Crackdown People of Internet Research · Taiwan 184 Disinformation investigations opened Police cases ahead of the 28 Novem… 112 Referred to judicial authorities About 61% of opened investigations… 4 of 9 Foreign-collusion cases charged Cases involving external forces, c… 0 Official deepfake takedown requests Central Election Commission count … peopleofinternet.com
Taiwan's Election Disinformation Crack… People of Internet Research · Taiwan 184 Disinformation investigations o… 112 Referred to judicial authori… 4 of 9 Foreign-collusion cases charged 0 Official deepfake takedown requests peopleofinternet.com

Key Takeaways

On 4 November 2026, Taiwan's National Police Agency said that as of the end of October it had opened 184 suspected disinformation investigations ahead of the 28 November local elections. It has referred 112 of them to judicial authorities. Separately, nine cases of suspected collusion between external forces, criminal groups and private detective firms are under investigation, and prosecutors have charged in four, according to the Taipei Times report. Candidates can pay the Criminal Investigation Bureau to forensically examine suspected deepfakes, and can report criminal content to police and to platforms. President Lai Ching-te also warned that many suspicious accounts with overseas phone numbers but Taiwan locations had appeared on social media in a short period, with consistent activity patterns.

The notable feature of this crackdown is the legal tool being used. Taiwan is relying on criminal and election law, not a dedicated platform statute. That choice is not accidental, and it carries real consequences for speech.

The strongest case for tougher action

The argument for aggressive enforcement is serious. Taiwan faces a sustained, state-linked influence effort. The National Security Bureau told reporters in October that China is using disinformation, illegal betting pools and hacking, and that it is gathering evidence to hand to prosecutors before polling day. The Taipei Times reported the NSB's description of local networks as mostly criminal gangs or private investigation firms. A hostile actor that pays locals to run accounts is committing something closer to organised crime than to protected speech. Election timelines are short, and a deepfake released days before a vote can do damage that a later correction cannot undo. On that view, the state should be able to act quickly and compel platforms to cooperate.

Why the platform route has not worked

Taiwan did try a platform-facing mechanism. A May 2023 amendment added Article 51-3 to the Public Officials Election and Recall Act. As described in Tech Policy Press, a candidate can obtain a police forensic finding that content is a deepfake and then demand that platforms remove it within two days, with fines set by the Central Election Commission. The author reports that the Commission had confirmed zero requests as of May 2026, so no election deepfake had been reported through the official channels.

The article gives several reasons. The Act protects only candidates, prospective candidates and recall parties, and only from the official election notice, typically 40 to 120 days before polling. Police findings are "notifications of fact" that cannot be appealed in court. And the enforcement target is hard to reach. According to the same piece, TikTok's entities are in Singapore, Taiwan Facebook Limited is registered for market research and advertising, and courts have held that legal requests to Meta belong with its US parent. A two-day removal clock cannot be enforced against a company that says it is outside your jurisdiction.

The Ministry of Digital Affairs has built a workaround. Its Election Deepfake Report site accepts reports only from candidates or prospective candidates themselves, who must authenticate with a natural-person certificate and upload a police forensic report. The ministry relays these to Meta, Google (YouTube), LINE, TikTok, PTT and Dcard. The page cites Article 51-3 as its legal basis, but it describes no obligation on platforms and no fines. Cooperation is voluntary.

The Ministry of Justice said in September that agencies would take down deepfake content promptly. The Taipei Times reported that spreading deepfake audio or video aimed at candidates can carry up to seven years in prison. I could not retrieve the statute text itself, so that figure rests on the newspaper's account.

What the criminal-law route gets right and wrong

There is a case that the current approach is the better one, and it should be stated plainly. Criminal and election law targets conduct: fraud, vote-buying, paid coordination, impersonation and defamation. These offences have evidentiary standards, defendants, and judges. Charging four of nine collusion cases is how a rule-of-law system handles paid interference. It does not require a regulator to decide what counts as "false" in the abstract.

A dedicated platform law would do the opposite. It would hand a government agency authority over what hundreds of millions of posts may say, enforced through fines on intermediaries that respond by over-removing. Tech Policy Press records that opposition lawmakers have already described the ministry's reporting site as a tool that turns officers into "political censors." That criticism is overstated, since the site is limited to candidate-initiated deepfake reports backed by forensic evidence. But it shows how quickly trust erodes when a government that is a political party's rival also runs the takedown pipeline.

The risks on the criminal side are different but real. Police opened 184 investigations and referred 112, which is 61 percent. That is a high referral rate for a category that includes satire, error and heated opinion. Referral is not conviction, and the public figures do not say how many referrals end in charges, or how many involve a candidate's reputation rather than foreign direction. Publishing that outcome data after the election should be a minimum expectation.

What proportionate policy looks like

Three measures follow from the evidence, and none requires a speech-licensing regime.

The broader lesson applies beyond Taiwan. An EFF analysis published in October argues that automated moderation errors around elections are costly and that platforms should be transparent and keep humans in the loop. If platforms are asked to act within two days, they will lean on automation, and mistakes will fall on legitimate political speech.

Taiwan's police are doing what a democracy should do against paid, foreign-directed manipulation: investigating conduct and taking it to prosecutors. The weakness lies in the unused platform mechanism and the unanswered question of who the law can actually compel. A domestic-representative rule would fix that more cleanly than any broader falsehood law.

Sources & Citations

  1. Taipei Times: NPA cracks down on election meddling (5 Nov 2026)
  2. Ministry of Digital Affairs: Election Deepfake Report site
  3. Focus Taiwan (CNA): NSB on China's targeted election interference (22 Oct 2026)
  4. Tech Policy Press: Taiwan's Election Needs a Deepfake Law That Works
  5. Taipei Times: Taiwan to crack down on deepfakes ahead of voting (21 Sep 2026)
  6. EFF: Social Media When Elections Are Close