The Model Artificial Intelligence Law (MAIL) v.3.0 - Multilingual Version
Creators
- ZHOU, Hui
- CAO, Yanlin
- CHEN, Tianhao
- DI, Xingsi
- FENG, Zixuan
- FU, Hongyu
- HE, Bo
- HONG, Yanqing
- HU, Naying
- LI, Xueyao
- LIN, Beizheng
- LIU, Canhua
- LIU, Jinrui
- MA, Lan
- SU, Yu
- WANG, Jun
- WANG, Lei
-
WANG, Wei
- WANG, Xiang
- WANG, Yue
- WU, Lance Fan
- XU, Jiujiu
- YAO, Zhiwei
- YUAN, Kang
- ZHANG, Song
- ZHANG, Xin
- ZHANG, Yan
- ZHAO, Shuyu
- ZHU, Yue
- ZHU, Lingfeng
Contributors
Others:
Description
The Model Artificial Intelligence Law 3.0 (MAIL 3.0) constitutes the second iteration of China's first systematic legislative prototype for AI governance. This version reflects a maturing regulatory vision that integrates foundational principles of AI law, a risk-calibrated licensing and registry regime based on a negative list model, and a delineation of developer and provider obligations accompanied by liability and enforcement provisions. It marks an experimentalist shift from sectoral or reactive regulation toward a more comprehensive and anticipatory legal architecture.
Beyond articulating substantive and procedural duties, the draft advances an institutional design agenda that foregrounds regulatory adaptability, including the establishment of technically capable supervisory bodies, the operationalization of ethical review mechanisms throughout the AI lifecycle, and the integration of open-source governance and intellectual property considerations into the regulatory framework. Notably, it introduces a "safe harbor" mechanism for generative AI liability, reflecting an effort to balance innovation with accountability—particularly in contexts of uncertain causation or diffuse agency.
The MAIL 3.0 also places particular emphasis on structural support for small and medium-sized enterprises, through tax incentives, third-party evaluation systems, and open-source incentives. It introduces novel legal constructs such as whistleblower protections and end-user rights clauses—measures that signal an emerging concern with procedural justice and distributive equity in the governance of algorithmic systems. A policy adaptability assessment mechanism is also proposed, requiring that new or existing legal instruments be evaluated for compatibility with the evolving technological and safety/security landscape posed by AI.
Taken together, the Model Law offers not merely a regulatory template but a jurisgenerative intervention in global AI governance. It invites comparative reflection on the institutional preconditions of effective regulation, the normative vocabulary appropriate to AI-specific risks, and the role of domestic legislative experimentation in shaping transnational regulatory paradigms.
Files
*人工智能示范法3.0【英文版】 - THE MODEL ARTIFICAL INTELLEGENCE LAW (MAIL)_v.3.0_EN.pdf
Additional details
Additional titles
- Translated title
- 人工智能示范法3.0
- Translated title
- LOI TYPE SUR L'INTELLIGENCE ARTIFICIELLE (MAIL) v.3.0
- Translated title
- LEY MODELO DE INTELIGENCIA ARTIFICIAL (MAIL) v.3.0
Dates
- Issued
-
2025-06-28