Chinese Scholars Release 99-Article Draft AI Law With Four-Party Liability Framework
Summary
- • Zhejiang University scholars release 99-article draft AI law covering developers, providers, deployers, and users
- • Draft requires human final judgment in high-sensitivity AI decisions affecting life, health, or freedom
- • Mandatory AI damage liability insurance and national relief fund proposed for high-impact systems
- • China's national AI law remains preparatory; no official draft yet submitted to the NPC
Details
99-article draft law released by Zhejiang University
Released September 24, 2026 as 'Zhejiang University 1.0 version'; structured in 8 chapters covering the full AI lifecycle from development through deployment.
Four-party liability framework across the AI lifecycle
Assigns legal responsibility to developers, providers, deployers, and users based on their actual control over AI systems at each stage—not simply who created the system.
Mandatory AI damage liability insurance for high-impact systems
High-impact AI systems would require mandatory insurance; a national AI damage relief fund compensates victims when liability is unclear or the responsible party cannot pay.
Human oversight mandatory in high-stakes AI decisions
Mandates human final judgment in decisions affecting life, health, freedom, property, or opportunities—covering medical diagnosis, autonomous driving, judicial enforcement, and brain-computer interfaces.
Open-source model disclosure rules with limited liability shield
Requires documentation and risk disclosure for open-source AI models while offering limited liability protections to open-source developers, balancing transparency with innovation incentives.
China's national AI law still a preparatory NPC agenda item
Comprehensive national AI legislation has no official draft yet submitted for NPC review; this scholarly proposal signals the likely legislative direction policymakers may follow.
Second major scholarly AI law proposal; first was CUPL's 2024 draft
An earlier AI law proposal was drafted in 2024 by scholars at China University of Political Science and Law; the Zhejiang University draft represents an evolution of that scholarly effort.
Details compiled via Grok live web research from MLex and Chinese media
Structural details, chapter counts, and liability mechanisms sourced from MLex reporting and Sohu/Cover News coverage of the Zhejiang University proposal.
Policy = regulatory/legislative detail; Context = background for interpretation; Research = sourcing note
What This Means
A prominent Chinese university legal team has released a detailed 99-article draft AI law that would hold developers, providers, deployers, and users liable for AI-caused harm based on their actual control of the systems. The proposal's mandatory insurance and national relief fund mechanisms represent novel approaches to compensating AI victims not seen in other major jurisdictions. While scholarly proposals lack binding force, they frequently shape China's formal legislative process—and with national AI legislation already on the NPC's preparatory agenda, this draft signals the direction Chinese policymakers are likely to take. Global AI companies operating in China should treat this as an early indicator of coming compliance requirements around liability, human oversight mandates, and open-source disclosure.