探索与争鸣 ›› 2024, Vol. 1 ›› Issue (7): 127-136.

• 技术与文明 • 上一篇    下一篇

行刑二元:生成式人工智能治理的新范式

刘宪权,董文凯
  

  • 出版日期:2024-07-20 发布日期:2024-08-09
  • 作者简介:刘宪权,华东政法大学功勋教授、“经天学者”讲席教授; 董文凯,华东政法大学刑事法学院博士研究生。(上海 201620)
  • 基金资助:
    国家社会科学基金重大项目“网络时代的社会治理与刑法体系的理论创新”(20&ZD199)

Dual Governance of Administrative Law and Criminal Law:A New Paradigm of Generative Artificial Intelligence Governance

Liu Xianquan & Dong Wenkai
  

  • Online:2024-07-20 Published:2024-08-09

摘要:

涉生成式人工智能违法行为的治理必然是一个由多主体参与、多环节监督、多部门法规制的多元共治的过程。其中,科技治理理论和法秩序统一原理是生成式人工智能行刑二元治理模式的理论依据;而源于违法一元论与违法相对论的立场之争,生成式人工智能的法律风险可能存在行刑界定之惑。为了减少技术革新对法律秩序稳定性的影响,我国应当首先在前置法层面制定规制涉生成式人工智能一般违法行为的行政法,然后再在刑法中增设规制涉生成式人工智能犯罪行为新类型的罪名,进而呈现出由行及刑(由行政法等前置法向刑法延伸)的立法结构。在行政法等前置法中,应当构建基于风险程度的分类分级管控制度,建立并完善生成式人工智能市场准入审查、全流程数据透明度披露以及事前合法性评估等机制。对于生成式人工智能等新兴领域的新型法定犯,可以将相关犯罪的构成要件以附属刑法的方式直接规定在行政法等前置法之中,同时将相关犯罪的法定刑统一规定于刑法中。

关键词:

Abstract:

The governance path for generative artificial intelligence technology risks must be a multi-disciplinary and co-governance process involving multi-subject participation, multi-link supervision, and multi-legal regulation. Generative artificial intelligence has risk characteristics such as systemic, technical, complex, and ethical risks. The theory of technological governance and the principle of unity of legal order are the theoretical basis for the dual governance model of generative artificial intelligence execution. Due to the dispute between illegal monism and illegal relativism, the risk of generative artificial intelligence technology may lead to confusion in the definition of execution. In order to reduce the impact of technological innovation on the stability of criminal law, we should first formulate an administrative law specifically regulating the risks of generative artificial intelligence technology at the level of preemptive laws. In preemptive laws such as administrative law, a risk-based classified and hierarchical control system should be constructed, and mechanisms such as generative artificial intelligence market access review, full-process data transparency disclosure, and ex-ante legality assessment should be established and improved. For new statutory crimes in emerging fields such as generative artificial intelligence, we should try to directly stipulate the relevant crime and punishment content in the preemptive laws in the form of subsidiary criminal law.

Key words: