作者：郭雪慧
作者介绍：中国民航大学法学院教授
出版日期：2026年10月9日
内容提要 人工智能证据在司法实践中的应用日益增多，但学界对其是否应成为独立的证据类型存在较大争议。人工智能证据是指，以具备自主数据处理能力的人工智能系统为核心生成主体，由算法驱动自主或半自主生成，承载能够证明案件事实的客观化信息，并服务于诉讼证明活动的证据材料。将人工智能证据归入书证、鉴定意见、电子数据等现有证据类型均存在理论与实践层面的困境。人工智能证据在生成主体、生成机制和表现形式等方面具有区别于传统证据的本体独特性，且对传统证据的准入规则、审查规则等形成系统性冲击。因而，从归类验证、本质剖析与规则研判三个层面分析，确立人工智能证据的独立证据地位具备理论依据。构建体系化的人工智能证据规则，应统筹修订三大诉讼法，将人工智能证据增设为独立证据类型；构建以合法性、相关性与可靠性为核心的证据准入标准，并设置附条件准入的例外规则；建立分类审查规则、人机协同纠错规则、可采性验证规则等证据审查规则。
关键词 人工智能证据 独立性 证据类型 证据准入 证据审查
Abstract: The application of artificial intelligence evidence in judicial practice is increasing, but there is a significant controversy in the academic community over whether it should be regarded as an independent type of evidence. Artificial intelligence evidence refers to evidence materials that are generated by an autonomous data processing system with artificial intelligence as the core, driven by algorithms to autonomously or semi-autonomously generate, carrying objectivized information capable of proving the facts of the case, and serving the litigation proof activities. Assigning artificial intelligence evidence to existing evidence types such as documentary evidence, expert opinions, and electronic data all have logical and practical application dilemmas. Artificial intelligence evidence has unique ontological characteristics in aspects such as generation subject, generation mechanism, and manifestation form that are different from traditional evidence, and it also forms systematic impacts on the admission rules, review rules, and other rules of traditional evidence. Therefore, analyzing from the three aspects of classification verification, essential analysis and rule evaluation, establishing the independent evidence status of artificial intelligence evidence has a theoretical basis. On this basis, artificial intelligence evidence should be added as an independent evidence type; establishing evidence admission standards centered on legality, relevance, and reliability, and setting an exception mechanism for conditional admission; establishing evidence review rules such as classified review and human-machine collaborative error correction, thereby constructing a systematic set of rules for artificial intelligence evidence.