On July 24, 2026, a short video titled "Oil Cannot Defeat You" (油不得你) began circulating online. It used AI-generated audio built to sound like President William Lai Ching-te (賴清德) criticizing his own government's handling of a food-safety crisis — batches of cooking oil from Central Union Oil Corp. that were recalled after testing found the carcinogen benzopyrene. The clip surfaced a day before the Kuomintang (KMT) held a rally on Ketagalan Boulevard over the same scandal. Taiwan's Criminal Investigation Bureau opened an inquiry after receiving public complaints on July 26, and identified the video's creator as Wei Chun-you, a former director of the KMT's New Media Department (Taipei Times; Mondaq).
A statute stretched past its design
Taiwan has no standalone deepfake law, so prosecutors are reaching for Criminal Code provisions written for paper documents: Article 210 (forgery of a private document), Article 216 (distributing a forged document), and Article 220 (treating audio/video as a "document" when it functions as proof of a statement). Whether those apply turns on whether the clip could plausibly be mistaken for something Lai actually said — a test built for forged signatures, now doing double duty against synthetic video. DPP spokesman Justin Wu argued the video "could constitute the crime of forging private documents"; law professor Chin Mong-hwa flagged additional exposure under the Personal Data Protection Act and personal-reputation and voice-rights claims, and called for mandatory labeling of AI-generated content (Taipei Times).
The KMT's counter is not frivolous: it has accused the government of selective enforcement and warned the investigation risks a chilling effect on political commentary, asking why a citizen was targeted while the oil-safety crisis itself went unaddressed. That tension — a law never built for synthetic media, applied to a case with an obvious political charge on both sides — is exactly why Taipei Times' editorial board asked plainly on August 3: "Is it time for a deepfake law?" (Taipei Times).
The case for going further
The strongest version of that argument isn't about this one video. It's about what Taiwan is exposed to. DPP Legislator Puma Shen has argued that deepfaking a sitting president during an active public-health controversy is high-risk activity that existing fraud and forgery law was never built to deter, and that Taiwan's exposure to China-linked information operations ahead of future elections makes a narrow, case-by-case forgery prosecution an inadequate response — you don't want the first real test of your legal system to be an ambiguous one. The Ministry of the Interior has taken a similarly hard line, with official Liu Shyh-fang stating that AI-generated impersonation of this kind isn't protected speech at all (Taipei Times). Given how thin the current toolkit is — reputation torts and a forgery statute doing work it wasn't drafted for — the instinct to legislate specifically for synthetic media is reasonable, and other democracies have already moved: the EU's AI Act mandates deepfake labeling, the US has the Take It Down Act, and Japan is extending voice-rights protection against AI cloning.
Why Taiwan doesn't need to start from scratch
The steelman still overstates the gap. Taiwan has spent the past two years building exactly the components a targeted response would need — it just hasn't assembled them for this case. The Fraud Crime Hazard Prevention Act, promulgated July 31, 2024, already forces online advertising platforms to disclose when an ad "uses deep fake technologies or AI-generated individual images," backed by fines of NT$1–25 million for serious violations and the power to order ISPs to restrict access to non-compliant platforms (Fraud Crime Hazard Prevention Act, Art. 40). The Public Officials Election and Recall Act separately criminalizes deepfakes used to disseminate false statements about candidates. And on top of both, the Legislative Yuan passed the Artificial Intelligence Basic Act on December 23, 2025 — Taiwan's first framework AI statute, built around seven governance principles including transparency and accountability, with the Ministry of Digital Affairs tasked with developing risk-tiered implementing rules (MODA press release).
What's missing isn't a legal category — it's a bridge between these three regimes and political speech about public officials, which today falls outside the ad-fraud law's scope and outside the election act's campaign-period window. The proportionate fix is to extend the disclosure model Taiwan already tested on advertising platforms — clear, narrow, technology-specific labeling requirements for AI-generated audio/video impersonating identifiable public figures — rather than draft an omnibus platform-liability statute modeled on the EU's Digital Services Act. Taiwan already tried that broader approach once: the NCC's 2022 Digital Intermediary Services Act draft collapsed after civil-society groups and ISPs revolted against its takedown and liability provisions. A narrow voice/likeness disclosure mandate, paired with clearer sentencing guidance for Articles 210, 216, and 220 when the "document" is synthetic media, would close the actual gap this case revealed without repeating that fight or handing regulators discretion to police political criticism dressed up as fraud prevention.
The cooking-oil video is a genuine warning sign — synthetic media impersonating a head of state during a live public-health scandal is not a hypothetical harm. But the answer is to finish stitching together the AI Basic Act, the Fraud Crime Hazard Prevention Act, and the election law's deepfake provisions into a coherent disclosure regime for public figures, not to write a new standalone statute whose scope will inevitably become the next partisan battleground.