Source: Legal Theory Blog

The article presents an in-depth conceptual, normative, and doctrinal examination of tort liability for the development and release of foundation models, arguing for a qualified defense of the tort of negligence as the principal doctrinal foundation for governing risks posed by such AI systems. It contends that negligence’s generality and flexibility, together with its capacity to account for externalized benefits of risky activities, make it uniquely suited to address multiple pathways by which foundation model developers might cause serious harm, including internal deployments, inadequate model-weight security, targeted entrustments of non-defective models, and open-source releases.

The author recommends specific incremental developments in common-law negligence doctrine—such as expanding the scope of the duty of care to cover economic or emotional injuries caused by model behavior that would violate human laws or norms, and recognizing a malfunction doctrine enabling plaintiffs to reach juries without additional breach evidence. At the same time, the article identifies structural pathologies of tort liability: torts may disincentivize frontier developers from investigating and disclosing novel risks, potentially worsening societal ability to discover such hazards. The piece concludes that tort law is inadequate in some respects and that a robust regime of ex ante regulation, with government institutions or neutral third-party experts empowered to investigate and mitigate AI risks, is urgently required. The blog post endorses the paper as timely and important and recommends downloading it.