Taiwan Investigates AI-Cloned Presidential Voice, Testing the Limits of Forgery Law

Attribution: 中华民国总统府

Taiwan's Criminal Investigation Bureau has opened an investigation into a video that allegedly used an AI-generated voice clone of President Lai Ching-te, in a case that is now testing how far existing forgery law can stretch to cover synthetic political speech.

The video, titled "Oil Cannot Defeat You," was published by a creator known as "Principal Lai'er," identified as Wei Chun-you, a former director of the opposition Kuomintang's New Media Department. The Criminal Investigation Bureau opened its inquiry on 26 July after receiving public complaints that the video used AI-generated audio to imitate the president's voice in a way that could mislead viewers into believing he had made the statements himself.

Taiwan has no dedicated statute for AI voice cloning, so legal experts have pointed instead to the island's existing forgery provisions. Under Article 220 of the Criminal Code, audio or video that functions as "proof" of a statement can be treated as a document. Fabricating such a recording may fall under Article 210, forgery of a private document, while distributing it may separately engage Article 216. Legal commentators have stressed that deepfake technology is not inherently illegal; liability turns on whether the content impersonates a real person, has been fabricated or altered, and is capable of causing harm.

The case has split opinion along political lines. Government legislators have argued that free speech does not extend to fabrication or defamation, and note that forgery is a public prosecution offence in Taiwan, meaning the state can act without the victim filing a personal complaint. Opposition figures have questioned whether the investigation itself risks chilling legitimate speech. With local elections due at year-end, several lawmakers have warned that cheap, high-volume AI-generated content could make it harder for voters to distinguish genuine statements from fabricated ones.

For rights holders and public figures across the region, the case is a useful marker of how differently APAC jurisdictions are approaching the same problem. China has introduced mandatory labelling and encrypted watermarking requirements for deepfakes, alongside a ban on watermark-removal tools. South Korea has moved to increase penalties specifically for creating or distributing illegal deepfakes. Taiwan, by contrast, is applying a decades-old forgery statute to a technology it was never written for. None of the three approaches yet offers a settled evidentiary standard, which means a synthetic voice or likeness that meets a criminal threshold in one jurisdiction may only trigger a labelling requirement, or no obligation at all, in another.

References

1. Central News Agency (CNA), reporting that Taiwan's Criminal Investigation Bureau opened an investigation on 26 July 2026 into the AI-voice video "Oil Cannot Defeat You," 27 July 2026. https://www.cna.com.tw/news/aipl/202607270077.aspx

2. Taipei Times, on the video's creator, "Principal Lai'er" (Wei Chun-you), a former director of the Kuomintang's New Media Department, 28 July 2026. https://www.taipeitimes.com/News/taiwan/archives/2026/07/28/2003861507

3. CNA, legal analysis of Articles 210, 216 and 220 of Taiwan's Criminal Code and the treatment of audio/video as "documents," 26 July 2026. https://www.cna.com.tw/news/aipl/202607260088.aspx

4. Vision Times, "AI-Cloned Voice of Taiwan President Lai Ching-te Sparks Debate on Free Speech and Deepfake Crimes," 28 July 2026. https://www.visiontimes.com/2026/07/28/ai-cloned-voice-of-taiwan-president-lai-ching-te-sparks-debate-on-free-speech-and-deepfake-crimes.html