隨區塊鏈技術之快速發展，⼈⼯智慧（AI）已逐步導⼊虛擬資產監理領域 ，成為防制洗錢與合規監控之核⼼⼯具， 惟當前監理科技 （RegTech）多採⽤具備⾼度複雜性之機器學習模型，其「⿊盒⼦」特性導致決策邏輯難以被外界理解，進⽽對⾏政程序正當性、資訊隱私權及受處分⼈之救濟權利形成嚴峻挑戰。國際間對於 AI 治理已趨向強調「透明性」 與 「可解釋性」，以確保⾃動化決策過程之公平性與可問責性。
歐盟、英國及美國等司法管轄區在 AI 監理法制⽅⾯已取得顯著進展。歐盟透過 《⼀般資料保護規則》 （GDPR）及 《⼈⼯智慧法》 （EU AI Act）確⽴了資料主體之解釋權；英國資訊專員辦公室（ICO） 則針對 AI 決策透明化提出具體之解釋模式；美國則側重於功能性監理，確保⾃動化⼯具符合現⾏⾦融合規標準。
本論⽂旨在通過分析國內外 AI 監理法制之現狀，探討適合我國虛擬資產監理之可解釋性標準，透過引⼊「主體中⼼」之解釋路徑，並建議於未來相關法令中納⼊「反事實解釋」及「⼈為介⼊」機制，期能於提升監理效率之同時，完善對虛擬資產使⽤者之法律保障，並促進我國數位⾦融產業於法治基礎上之健全發展。;The rapid advancement of blockchain has integrated Artificial Intelligence (AI) into virtual asset supervision as a core tool for anti-money laundering (AML). However， the "black-box" nature of advanced Regulatory Technology (RegTech) creates decision-making opacity， posing challenges to administrative due process and the remedial rights of affected parties. Consequently， international AI governance is shifting toward prioritizing "transparency" and "explainability" to ensure fairness and accountability.
While jurisdictions like the EU and UK have established legal frameworks， such as the GDPR and EU AI Act to secure data subjects′ right to explanation， Taiwan still faces regulatory gaps. Although Taiwan has introduced the "Artificial Intelligence Basic Act，" specific protections against automated decisions in virtual asset supervision remain insufficient.
The heavy reliance of Virtual Asset Service Providers (VASPs) on third-party tools further exacerbates information asymmetry and evidentiary hurdles for users.
This thesis analyzes domestic and international AI regulations to propose explainability standards tailored for Taiwan′s virtual asset supervision. By advocating for a "subject-centric" approach and incorporating mechanisms like "counterfactual explanations" and "human-in-the-loop，" this research aims to balance supervisory efficiency with legal protection， fostering a robust legal foundation for the digital finance industry.
