The AI Liability Puzzle and A Fund-Based Work-Around

The current legal system struggles with AI harms that are unforeseeable: The paper identifies a critical gap where existing liability laws assume harms can be predicted, but some AI systems may cause damage in ways that courts cannot reasonably have anticipated, creating legal uncertainty that actually discourages responsible AI development rather than promoting it.

Unpredictable liability makes AI investment too risky: Because companies cannot calculate the legal risks of deploying AI systems—courts might rule differently in similar cases involving unforeseeable harms—investors and developers face "incalculable" risks that may prevent beneficial AI applications from reaching society.

A fund-based compensation system could replace traditional lawsuits for AI harms: Drawing from financial regulation models, the authors propose "AI guarantee schemes"—dedicated funds that would compensate victims of AI-caused harm without requiring them to prove foreseeability through conventional litigation, similar to how insurance pools work in other industries.

Industry funding could align incentives with safety: The proposed system would be at least partially funded by AI companies themselves, which could encourage manufacturers to invest in safety measures and responsible practices, since they would bear some financial responsibility for harms their systems cause.

Clear rules would benefit both innovation and public protection: By establishing transparent, predictable compensation mechanisms upfront, society could enable AI development to proceed with confidence while ensuring victims have a reliable path to compensation, rather than leaving both innovators and the public in legal limbo.