重思深度偽造影音的入罪化──以人工智慧的科技風險規制為思考方向

Reconsidering on the Criminalization of Deepfakes - A Perspective on the Regulation of Technical Risks from AI

近來深度偽造的議題因為網紅小玉案的一審判決再度發酵，各方對於法院依據現行刑事法所作成的判決再度表示不滿。在此之前，立法者早已加快了立法的腳步，增訂相關的新法加以因應。然而，深度偽造的入罪化並不僅涉及到單一法益（如人格權）的保護，更涉及深度偽造實際可能引發的各類風險評估、科技風險倫理的立場、刑法現存規定的適用可能性與人工智慧技術應用對未來整體社會或個人的影響層面等問題，其解決方案並非單靠一條刑法條文就一蹴可幾。本文就此除了簡要評介2023年新法之外，更從正反面的不同影響評估深度偽造的潛在風險，並指出現行刑法的適用可能性，最終提出本文對於深度偽造入罪化的立場：就個人法益的保護而言，現行法已經有誹謗罪與個人資料保護法等規定得以適用，並非完全無法可用，新增保護人格權法益的犯罪類型並非絕對必要。就集體法益的保護而言，則有待先對人工智慧技術建立起相關體制與具體法規範，後續才有刑法介入的空間。就此而言，立法者所提的新法內容恐只有事倍功半的效果。

“Deep fake” is currently a remarkable phenomenon in the digital world. Also the actions of producing and distributing images, sounds as well as movies by means of the technique of deepfakes are lege ferenda to think whether it should be punishable. Due to a scandal from a Youtuber, the government in Taiwan has come up with a draft to fight against deep fakes in 2022. However, the draft is very questionable because the elements of the offense and the legal interest to be protected remain highly unclear in the provision. The author believes that the problem of deep fakes can already be partially solved with the current regulations in criminal law. If a new offense to fight against deep fakes is absolutely necessary, the legislator should not neglect to include the misuse of the technology of AI.