司法通譯與人工智慧協作──應用潛力與制度挑戰

Judicial Interpreting and Artificial Intelligence Collaboration:Application Potential and Institutional Challenges

隨著人工智慧（AI）技術的迅速發展，語音辨識與自然語言處理工具已逐步導入司法程序，為司法通譯提供新的輔助契機。本文以臺灣司法通譯制度為研究核心，探討AI導入通譯工作的可行性、制度挑戰與倫理風險。研究首先從現行制度層面出發，分析司法通譯在人力資源不足、語種不均、培訓體系不健全與品質監督缺口等結構性困境，並指出該制度在面對多語社會與新住民需求時的侷限。其次，本文結合《人工智慧基本法（草案）》之四項核心原則——第3條「公平與透明」、第10條「可驗證性與人為可控性」、第11條「高風險AI應用」與第13條「個人資料保護」——從法律與倫理雙重視角，評估AI協作在司法通譯制度中之應用潛能與治理風險。
研究後半部以「泰國籍移工酒駕案件」之司法模擬訓練為案例，提出「人機協作智慧通譯架構」，展示AI於審理前、審理中與審理後三階段的輔助功能：語料準備、語音轉錄與逐字稿檢核。結果顯示，AI技術可有效提升通譯效率與記錄精準度，並強化語料追蹤與責任釐清；惟若缺乏責任歸屬、資料保護與倫理審查機制，將可能導致語境誤判、偏誤傳譯及當事人語言權受損等新型風險。本文最後提出五項政策建議，包括建立AI司法應用倫理指引、技術審查與認證制度、責任鏈設計、資料匿名化與隱私保護，以及建構人機協作通譯制度。
本研究延續筆者《法庭口譯：理論與實踐》之人文法學觀點，主張AI應作為通譯專業的「協力者」（collaborator）而非「取代者」（substitute），並指出唯有在法制、技術與人權並進之下，方能建構兼顧效率與程序正義的智慧司法語言服務體系，實現「科技促進正義」之願景。

With the rapid advancement of artificial intelligence (AI), technologies such as automatic speech recognition (ASR) and natural language processing (NLP) have been gradually introduced into judicial proceedings, offering new opportunities to assist court interpreters. Centering on Taiwan’s judicial interpreting system, this study investigates the feasibility, institutional challenges, and ethical implications of integrating AI into interpreting practices. The research first analyzes structural deficiencies—such as interpreter shortages, uneven language distribution, inadequate training systems, and insufficient quality supervision—and identifies the institutional limitations of Taiwan’s current framework in addressing multilingualism and the linguistic needs of new residents. From both legal and ethical perspectives, the study applies four key principles of the Artificial Intelligence Fundamental Act—Article 3“Fairness and Transparency,”Article 10“Verifiability and Human Controllability,”Article 11“High-Risk AI Applications,”and Article 13“Personal Data Protection”—to evaluate the potential and risks of AI collaboration in judicial interpreting.
Using a simulated DUI case involving a Thai migrant worker, the study develops a Human–AI Collaborative Interpreting Framework, illustrating AI’s supportive functions in three procedural stages: pre-trial (corpus preparation), in-trial (speech transcription), and post-trial (verbatim review). Findings reveal that AI can enhance interpreting efficiency, transcript accuracy, and traceability while reducing workload pressures. However, without robust accountability, data protection, and ethical oversight, AI integration may introduce new risks, including semantic misjudgment, biased translation, and the erosion of linguistic rights.
The paper concludes with five policy recommendations: establishing ethical guidelines for AI judicial applications, implementing certification and review mechanisms, defining liability chains, enforcing data anonymization and privacy safeguards, and institutionalizing a human–AI collaborative interpreting system. Building upon the author’s earlier work Court Interpreting: Theory and Practice (2018), this study contends that AI should act as a collaborator rather than a substitute for human interpreters. Ultimately, it argues that only through the integrated advancement of law, technology, and human rights can Taiwan construct an intelligent judicial language service system that balances efficiency with procedural justice, advancing a human-centered model of justice through.