Tort Law at the Frontier of Artificial Intelligence

The frontier of contemporary AI development is dominated by AI systems built on foundation models—highly versatile algorithms, trained in the first instance on broad swaths of data, that can function as speakers, tools, and agents across a wide variety of commercial, social, military, and political domains. For the moment, at least, the process of developing and releasing foundation models is subject to anemic ex ante regulation. Until that changes, it is largely the common law of torts—our society’s most general legal mechanism for governing serious risks of physical injury—that will govern the frontier of AI development.  

This Article offers an in-depth conceptual, doctrinal, and normative examination of tort liability for foundation-model development and release. It provides a qualified defense of the tort of negligence, the common law’s most general and flexible cause of action, as the principal doctrinal foundation of the tort system’s governance of this novel domain. Legal scholarship on AI liability—much of which predates the advent of foundation models—has been largely hostile to negligence, arguing that AI should principally or exclusively be governed by alternative doctrinal regimes. By contrast, this Article argues that the generality and flexibility of the negligence tort—and its greater sensitivity to the externalized benefits of risky activity—render it better suited to the polymathic and protean character of foundation models and the complex technical and institutional dynamics of their development and release.  

Analyzing the choice between negligence and competing doctrinal regimes does, however, reveal important ways in which the courts should incrementally develop the law of negligence in order to properly address the nature and risks of foundation models. First, courts should expand the scope of the duty of care in negligence in order to provide redress when foundation models cause economic or emotional injury by behaving in a manner closely analogous to serious human wrongdoing. Second, courts should import into the law of negligence a malfunction doctrine, one similar to that which already exists in products liability. When a model behaves in a manner that is closely analogous to an intentional tort or crime, or that clearly contravenes legitimate expectations of safe performance, courts should allow the plaintiff to reach the jury even without adducing (further) evidence on breach.

The Article’s analysis also suggests certain limitations and pathologies of tort liability as a mechanism of AI governance—issues that cannot be adequately addressed by any amount of judicial doctrinal development or legislative modification of the applicable liability rules. In particular, the specter of tort liability can be expected to disincentivize frontier AI developers from investigating and disclosing the novel and poorly understood risks that frontier- AI development may pose. That is especially disturbing given that our society is—unavoidably, to a large extent—relying quite heavily upon frontier AI developers themselves in order to discover, understand, and mitigate these risks. So, as a mechanism of frontier-AI governance, ex post liability alone is not only inadequate but in certain respects perverse. Ultimately, a robust regime of ex ante frontier-AI regulation—under which government institutions or credibly neutral third-party experts are empowered to ensure that AI developers are properly investigating and mitigating the risks of foundation-model development—is urgently required.