| 摘要: |
| 人工智能立法没有必然的路径。欧盟《人工智能法》的制定采取了“产品安全思路”，试图借助对产品安全立法的模仿、准用与调整来回应人工智能风险。产品安全思路的意义在于通过标准化进程来整合市场、通过锁定“最小成本规避者”来实现安全威慑，并通过公法规范的私法效力来激活产品责任体制。然而，人工智能的非标准化特征预示了标准化进程的困难，人工智能的多元使用场景揭示了义务分配的复杂性，信息成本高昂的事前标准也难以激活产品责任的运作。考虑到中欧之间迥异的市场结构、制度语境与产品安全思路自身的局限，我国并无采纳产品安全思路的必要。相反，我国应认真对待人工智能领域的不确定性，借助责任体制的信息生产功能为下一步法律动作提供事实依据。 |
| 关键词: 人工智能法 欧盟 产品安全 技术标准 |
| DOI： |
| 分类号: |
| 基金项目:国家社会科学基金重大项目“当代中国数字法学基本范畴体系研究”（项目批准号： 23&ZD154） |
|
| Product Safety Approach for AI Legislation： A Critical Reading of the EU AI Act |
|
He Zehao
|
|
|
| Abstract: |
| AI legislation does not have an inevitable path. The European Union's AI Act adopts a “product safety approach” in its legislative strategy, attempting to respond to the risks of AI through the imitation, application and adjustment of product safety legislation. The significance of the product safety approach is to integrate the market through the process of standardization, to achieve safety deterrence by locking the “Cheapest Cost Avoider”, and to activate the liability system through the public law. However, the non-standardized nature of AI indicates the difficulty of the standardization process, the diversity of AI use cases reveals the complexity of obligation allocation, and the information-costly ex ante standards do not provide an “antidote” to the liability regime. Given the significent differences in market structures and institutional contexts between China and the EU, as well as the limitations of the product safety approach itself, there is no need for China to adopt the product safety approach. Instead, we should take the uncertainty in the field of AI seriously and use the liability system's information production function to provide a factual basis for the next legal action. |
| Key words: AI Legislation, EU, Product Safety, Technical Standards |
