Author

Jiyu Zhang, Professor, Law School; Research Fellow, Law and Technology Institute, Renmin University of China

Yuyang Tian, PhD Candidate, Law School, Renmin University of China

Abstract

Currently there is no consensus either domestically or internationally on the definition of open-source models, Open-source models can be classified based on the scope of open-source content the degree of permissiveness of their open-source licenses and whether commercial licensing fees are charged. They can also be distinguished as general-purpose foundational models or domain-specific models as well as large models or small models, In utilizing of generative artificial intelligence applications the fault-based liability principle should be applied to open-source model providers and the standards for determining the duty of care should be developed by taking into account multiple factors, For providers of free foundational large models that have high social empowerment value but limited risk control capacity and high prevention costs a lower duty of care should be imposed. Conversely for open-source small models with low social utility high security and ethical risks and foreseeable potential for infringing use the providers should be deemed at fault. Finally, to better encourage foundational open-source large models to empower society governments should promote the establishment of multiple co-governance mechanisms involving industry associations and other entities, The duty of care of open-source model providers should be aligned with the governance measures that policymakers wish to encourage. By doing so it is possible to enhance risk prevention capabilities without unduly increasing the cost burden on open-source model providers thereby fostering the responsible and beneficial development of open-source models.

Keywords: open-source models, tort liability, duty of care, multi-factor analysis, multi-stakeholder governance