欧盟人工智能的实验主义治理路径及中国借鉴

A Study of the Experimentalist Governance Approach of the EU’s Artificial Intelligence and Its Lessons for China

欧盟《人工智能法》采用的实验主义治理路径,借由多主体试错学习快速形成经验,能够及时回应人工智能技术的最新发展。与理论先行的理性主义治理路径相比,实践先行的实验主义治理路径设想建立一种基于契约的合作监管模式,实现监管关系的创新性;倡导设置分级分类的实验性框架目标,为新事物留足发展空间,实现监管目标的创新性;提倡构建扁平化的网状监管结构,将实验信息迅速上升为治理策略,实现监管结构的创新性。面对人工智能等新兴技术的监管问题,中国政府也在最新的人工智能治理策略中采用实验主义治理路径,并于2024年建成国内首个实验性人工智能训练基地。同时,为了防止行业自我监管的情况出现,中国政府可以根据分级分类的理念,对实验主义治理路径的适用领域和适用场景进行合理布局。

The EU Artificial Intelligence Act adopts an experimentalist governance approach,utilizing multi-actor trial-and-error learning to rapidly gain experience and promptly respond to the latest developments in AI technology.Compared with the rationalist governance approach,which is theory-driven,the practice-driven experimentalist approach aims to establish a contract-based cooperative regulatory model to achieve innovation in regulatory relationships.It advocates setting experimental framework goals for categorization and classification to allow ample room for the development of new entities,thereby achieving innovation in regulatory objectives.It also promotes the construction of a flat,networked regulatory structure to allow experimental information to be quickly elevated to governance strategies,thus achieving innovation in regulatory structures.In response to regulatory challenges posed by emerging technologies such as artificial intelligence,the Chinese government has also adopted an experimentalist governance approach in its latest AI governance strategies.In 2024,it established the country’s first experimental AI training base.Additionally,to prevent industry self-regulation,the Chinese government can rationally plan the applicable domains and scenarios for the experimentalist governance approach based on the principle of classification and categorization.