治理研究 ›› 2025, Vol. 41 ›› Issue (2): 20-37.

• AI时代的治理变革专题 • 上一篇    下一篇

我国人工智能立法的制度创新与路径选择

杨延超   

  • 收稿日期:2024-12-10 出版日期:2025-03-15 发布日期:2025-05-09
  • 作者简介:杨延超,中国社会科学院法学研究所研究员,科技与法研究中心主任;中国社会科学院大学教授,博士生导师,博士后合作导师。

Institutional Innovation and Path Selection in China’s AI Legislation

Yang Yanchao   

  • Received:2024-12-10 Published:2025-03-15 Online:2025-05-09

摘要:

DeepSeek等人工智能技术特别是大模型技术的迅猛发展,正从根本上冲击现代法律体系的价值基础和规则逻辑。在法律价值层面,AI展现出的自主性、类人理性和复杂决策能力,动摇了以人类理性、自由意志和道德判断为核心的法律价值体系,挑战了正义、平等与自由的传统法律目标。在法律规则层面,AI突破了传统民事权利和责任体系的框架,催生数据权、算法权等新兴权利,同时引发跨领域权力合法性与实效性危机,要求重新界定权利、义务与责任的分配。为应对这一挑战,我国人工智能立法需在制度创新和路径选择上探索平衡。一方面,需重构权利主体和客体制度,明确AI的工具属性并在特定场景下赋予其拟制主体性;另一方面,需创新权利、义务和责任内容,突破传统“完全支配权”“静态义务”和“过错责任”模式,构建动态适应AI特性的法律框架。在立法路径上,应采取“统一立法+分别立法”的双轨模式,重点解决跨部门协同、领域规制和国际规则衔接问题,推进“嵌入式创新”与“独立性规制”的结合。

关键词: 人工智能, DeepSeek, AI立法, 制度创新, 以模制模

Abstract:

The rapid rise of AI technologies like DeepSeek challenges modern legal systems’ values and rules. AI autonomy and decision-making disrupt our current legal foundations because they are based on human reason and morality, questioning justice, equality, and freedom. AI has also exceeded traditional rights and liability frameworks by introducing data and algorithmic rights while complicating legitimacy and responsibility allocation. China’s AI legislation must innovate by redefining legal subjects. AI should be treated as a tool with limited abstract subjectivity. Rights, duties, and responsibilities must be updated beyond the traditional static, fault-based models to suit AI’s dynamics. A dual-track “unified + sector-specific” legislative approach should address coordination, regulation, and global alignment. This will blend embedded innovation with independent oversight. Legislation should also balance innovation and risk through progressive laws and sandbox testing, ensuring adaptability and supporting AI’s healthy growth.

Key words: artificial Intelligence, deepSeek, aI legislation, institutional innovation, model-based regulation

中图分类号: