Product Liability as a Model for UK AI Security

Gaurav Yadav, mentored by Peter Wills, researched product liability as a tool for AI safety.

Summary

Negligence, as a legal framework, will struggle to meet two 'core objectives': prompting frontier AI developers to internalise and mitigate the risks posed by their systems, and offering a robust avenue of redress for those harmed by them. This paper argues that a 'Frontier AI' Bill in the UK should implement a liability regime to supplant negligence law, giving stakeholders—such as courts, the public, and the developers themselves—greater clarity over where liability lies. To achieve this, the regime should adopt four mechanisms from product liability—duty to warn, consumer-expectation tests, developmental-risk defences, and supply-chain liability—as foundations for governing frontier AI developers.