ABSTRACT
The release of open-source AI foundation models has been advertised as ‘good for the world’. But free public access comes with significant risks of harm. The tacit assumption has been that the ‘open source’ label carries with it automatic immunity from legal liability. Accordingly, critics and skeptics of open-source AI have focused their attacks primarily on whether open-source AI properly qualifies as ‘open source’.
But the rule of liability for open-source AI should not depend on openness. This Essay explains that the notion of open-source exceptionalism draws upon a discredited theory of charitable immunity. Conventional tort law principles do not allow charitable actors to commit harm with impunity. Although donations of code or AI to the public domain exhibit altruistic intent and high societal benefit, those public benefits do not offset the ordinary policy justifications to require exercise of due care.
Choi, Bryan H, Open Source AI, Open Liability AI (January 1, 2026), University of Colorado Law Legal Studies Research Paper No 26-9.
Leave a Reply