Synthetic Harm and Legal Accountability: Structuring Tort Law Responses to Generative AI Systems

Generative artificial intelligence, Synthetic harm, Tort liability, Negligence, Causation, Multi-actor liability, Product liability, Algorithmic accountability, Data provenance

Generative AI has come a long way from its humble beginnings as something that was merely bandied about in casual conversation among friends and colleagues. Today, it is ubiquitous and has become an integral part of every digital platform that we use and interact with. The extent to which it has become an integral part of our lives is raises significant normative and legal concerns. While its proliferation is certainly having many positive effects on our lives, it is evident that there are many problems that come along with it, particularly in the context of tort law principles in the United States. While it is evident that there is an element of uncertainty that is associated with Generative AI and its outputs, can we really rethink and reapply some of the very old principles of tort law to this new and complex world? This paper is based on three major legal disruptions that have been caused by generative AI. First of all, the unpredictability of generative AI makes it hard to apply principles of negligence law because it is based on foreseeability and a quantifiable standard of care. Secondly, tort law is based on the concept of one person being liable for an injury. On the other hand, generative AI is based on complex systems that spread liability to developers, companies that use the technology, and the technology itself. Thirdly, there is difficulty in applying product liability law to generative AI because it is dynamic and is changing all the time through interactions with users and is thus considered to be either a product or a service. Furthermore, there is an overwhelming amount of information on platforms such as Widely used generative AI platforms and that is seeping into regular social media which is making it increasingly difficult to differentiate between what is real and what is artificial. This is posing a threat to legal principles that depend on clearly identifiable communicators and intent. Instead of seeking to fundamentally transform the law of torts, this paper proposes a more pragmatic approach to developing a parallel system that can be applied in conjunction with existing rules. This might include risk-based categorization, ex-ante duties, and ex-post accountability. The intention is to address the current issues while also being flexible to address future changes in generative AI.