The legal inspection of a collaborative governance system for artificial general intelligence with soft law and hard law cooperation in China

다국어 초록 (Multilingual Abstract)

The advent of artificial general intelligence has exacerbated the risk levels and governance complexities of issues such as privacy protection, intellectual property, and data security. Following established patterns, the governance of legal risks associated with artificial general intelligence can take two paths: soft law and hard law governance. However, both soft law and hard law governance have their limitations. In soft law governance, technological barriers result in insufficient public participation, rendering it less effective, while the effectiveness of soft law itself is also limited. In hard law governance, there is a lack of sufficient expertise, making regulations ineffective; risks related to data and algorithms are concealed, and accountability is unclear, leading to frequent contradictions and practical impediments in hard law. Therefore, it is necessary to reassess these paths and replace them with a more flexible paradigm that maintains a balance between safety and development to address the various risks posed by technological iterations. The soft law-hard law collaborative governance system thus emerges.
The soft law-hard law collaborative governance system combines the flexibility of soft law governance with the rigidity of hard law governance. Theoretically, it allows for the governance of artificial general intelligence to move beyond branch law and decentralized governance towards a collaborative, multi-law governance; practically, it utilizes a multi-faceted governance based on technology, ethics, and systems, integrating government governance, judicial governance, and industry self-regulation in a flexible adjustment mechanism that both encourages the standardized development of technology and prevents its disorderly expansion. However, the effective operation of the soft law-hard law collaborative governance system relies on the clarification of cognitive premises, ontological definitions, and method applications.
Regarding cognitive premises, the regulatory nature of soft law should be strengthened, and the operability of hard law should be enhanced, with organic coordination between the two. Regarding ontological definitions, data and algorithms, as objects of governance for the risks of artificial general intelligence, are merely representations; the behaviors of participant as objects of governance are the essence. Regarding method applications, this includes the indirect application of soft law under the dominance of hard law and the direct application of soft law in the absence of hard law.