生成式人工智慧在刑事偵查決策輔助中的應用與合法性研究
陳文博、鍾雲東

摘要：
生成式人工智慧（GAI）在刑事偵查決策輔助中展現出重要潛力，特別是在信息整合、證據分析、線索研判及文書處理等關鍵環節，能夠顯著提升偵查效率與準確性。然而，其固有的“幻覺”現象、運行機制的不透明性、潛藏的算法偏見，以及數據安全與濫用風險，均對偵查活動的公正性與合法性構成嚴峻挑戰。現行法律框架在證據採信、程序保障與數據合規方面存在明顯不足，難以有效規範GAI的深度應用。鑑於此，有必要構建涵蓋技術標準、法律規制、程序保障、倫理監督與跨部門協同的多維治理框架，並透過試點探索與動態調整，實現偵查效能提升與公民權利保障之間的平衡，為GAI在刑事偵查領域的安全、有序發展提供制度與實踐支撐。

關鍵詞：
生成式人工智慧 刑事偵查 法律規制 風險治理 程序保障

The Application and Legality of Generative Artificial Intelligence in Criminal Investigative Decision-Making
Chen Wenbo; Zhong Yundong

Abstract：
Generative Artificial Intelligence (GAI) demonstrates significant potential in assisting criminal investigation decision-making, particularly in critical areas such as information integration, evidence analysis, lead assessment, and document processing, where it can substantially enhance investigative efficiency and accuracy. However, its inherent “hallucination” phenomenon, opaque operational mechanisms, latent algorithmic biases, and risks of data security breaches and misuse pose severe challenges to the fairness and legality of investigative activities. The current legal framework exhibits significant shortcomings in evidence admissibility, procedural safeguards, and data compliance, making it difficult to effectively regulate the in-depth application of GAI. Given this, it is necessary to establish a multidimensional governance framework encompassing technical standards, legal regulations, procedural safeguards, ethical oversight, and cross-departmental coordination. Through pilot exploration and dynamic adjustments, this framework should balance investigative efficiency with the protection of citizens’rights, providing institutional and practical support for the safe and orderly development of GAI in criminal investigations.

Key words：
Generative Artificial Intelligence Criminal I nvestigation Legal Regulation Risk Governance Procedural Safeguard