A voice that was never his
On July 27, 2026, Democratic Progressive Party (DPP) spokesman Justin Wu told reporters in Taipei that prosecutors should open a full criminal investigation into an AI-generated video, "It's Out of Your Control" (油不得你), that used a synthetic clone of President William Lai Ching-te's voice to attack the government's handling of the Central Union Oil Corp. cooking-oil contamination scandal (Taipei Times). The Criminal Investigation Bureau confirmed it had received the complaint and was examining a video producer surnamed Wei. Wu argued the harm compounds when the impersonated figure is a head of state: "The societal impact is even more severe when government officials, the president or the vice president are targeted, necessitating a thorough judicial investigation."
The underlying scandal is real and serious. Central Union Oil was fined NT$165.2 million (~US$5.2 million) on July 7 — the largest fine a central government agency has levied in a Taiwanese food-safety case — after regulators found soy-based oil supplied to major processors between April and June carrying roughly four times the legal limit of benzo[a]pyrene, a carcinogen, and after the company delayed reporting the contamination and gave investigators inaccurate information (Focus Taiwan). More than a thousand tonnes of product were recalled. Public anger at the government's response is not manufactured; the deepfake exploited a genuine grievance.
The steelman: impersonating a head of state isn't ordinary satire
The DPP's case deserves to be stated fairly before it's contested. A synthetic clone of a sitting president's voice, packaged to sound like an official statement, is categorically different from a cartoon, a meme, or an actor's impression — the entire design goal is to borrow the authority of the office to make people believe the president said something he didn't. Legal scholar Chin Mong-hwa, cited in the same Taipei Times report, noted the video could implicate forgery, offenses against reputation, and Personal Data Protection Act (PDPA) violations. That's not a stretch: Taiwan's PDPA criminalizes the unlawful collection or processing of personal data — which extends to biometric identifiers such as a voice print — with penalties up to five years' imprisonment and a fine of up to NT$1 million for intentional violations (Laws & Regulations Database). Taiwan's Criminal Code separately criminalizes forging documents and causing them to circulate as genuine (Laws & Regulations Database). If a voice clone is engineered to pass as an actual presidential statement, applying forgery and privacy law to the impersonation itself is a defensible, narrow response — one grounded in what the content falsely claims to be, not in whether its message is politically inconvenient.
Why the KMT's alarm still has a point
That said, the opposition's objection isn't just noise. KMT caucus secretary-general Lin Pei-hsiang asked pointedly whether, "In the eyes of the government, is the president's ego more important than the public's health?" and accused authorities of turning enforcement into a chilling instrument — using the manner of the criticism as a pretext to investigate its substance (Taipei Times). The risk is structural, not hypothetical: any prosecutor deciding whether a synthetic-voice satire is "impersonation" or "commentary" is also, inevitably, adjudicating how sharply a citizen may mock the head of state. Selective enforcement — investigating deepfakes that embarrass the government while ignoring ones that don't — would turn a legitimate forgery statute into a partisan speech filter. Taiwan's prosecutors need to show, transparently, that the same standard would apply to a clone of an opposition leader's voice used the same way.
The road not taken
What makes this episode instructive is the law Taiwan is not using. In 2022 the National Communications Commission floated a Digital Intermediary Services Act modeled on the EU's Digital Services Act, which would have handed regulators broad authority to flag platform content. After heavy public backlash over speech concerns, the NCC halted drafting in September 2022 and, as of its own account, has set no timeline to revive it (Focus Taiwan). That restraint has held for four years, through an election cycle and now a genuine deepfake controversy. Taiwan is instead reaching for general-purpose forgery and privacy law — tools that require prosecutors to prove specific, individualized harm (a fabricated identity, misappropriated biometric data) rather than giving a regulator standing discretion to police "disinformation" as a category.
The proportionate path
The right test isn't whether AI-generated criticism of officials should ever be prosecutable — impersonating a president to fabricate an official statement crosses a line satire doesn't need to cross. The right test is whether the law used is narrow, applies identically regardless of whose ox is gored, and stops well short of a standing content-flagging regime. Taiwan's forgery-and-privacy approach clears that bar so far; a revived platform-liability statute would not. The DPP should get its investigation. It should also make certain, visibly, that the standard survives being pointed the other way — and that Taipei never needs to dust off the 2022 draft to answer the next deepfake.