摘要
大模型具有自主演化性、数据依赖性与算法黑箱性三大核心技术特征,由此引发输出内容安全性隐忧、数据使用正当性质疑、因果关系认定难题等风险挑战。大模型运行引发的损害不仅样态繁多,且具有独特的法律特征,传统归责体系面临前所未有的挑战,主要法域相关立法变革也存在缺陷。为此,需要借助风险分配与矫正正义的耦合来重构大模型运行损害责任分担的理论基础,创建兼容技术、控制和收益的三维归因模型。在此基础上构建多元主体的分层义务,遵循相应归责原则,利用原因力分析来合理划分比例责任,并结合动态风险协议制度来形成适应大模型技术特征与发展需求的动态、灵活且公正的责任分担机制。
Large models possess three core technical characteristics:autonomous evolution,data dependence,and algorithmic black-box nature,which give rise to risk challenges such as safety concerns regarding output content,legitimacy risks in data use,and difficulties in deter-mining causation.Damages arising from large model operations are not only diverse in form but also possess unique legal characteristics,posing unprecedented challenges to traditional attribu-tion systems.Furthermore,related legislative reforms in major jurisdictions also exhibit shortcom-ings.To address this,it is necessary to reconstruct the theoretical basis for liability sharing of damages caused by large model operations through the coupling of risk allocation and corrective justice,creating a three-dimensional attribution model compatible with technology,control,and benefit.Building upon this,a layered system of obligations for multiple stakeholders should be established,adhering to corresponding attribution principles,utilizing causality analysis to rea-sonably apportion proportional liability,and integrating a dynamic risk agreement system to form a flexible,dynamic,and just liability sharing mechanism adapted to the technical characteristics and developmental needs of large models.
出处
《东方法学》
北大核心
2025年第4期124-137,共14页
Oriental Law
基金
国家社科基金后期资助项目“国家出资公司中国特色治理机制研究”(项目批准号:23FFXB040)的阶段性研究成果。
关键词
大模型
人工智能
原因力
比例原则
损害责任
动态分担机制
large models
artificial intelligence
causality
proportionality principle
liability for damages
dynamic apportionment mechanism
