Trent Kannegieter (Yale U Law) has posted “Nondeterministic Torts: A Mechanistic Approach to Large Language Model Tort Liability” on SSRN. Here is the abstract:

Our laws were built for deterministic machines. When given the same inputs, we expect a system to consistently produce the same outputs. But modern artificial intelligence (AI) systems, specifically large language models (LLMs) and other types of “generative AI” (GenAI), challenge this assumption. These systems are nondeterministic, meaning that they produce varied outputs even when given identical inputs. By their nature, nondeterministic LLM-based predictions carry an arbitrary randomness that is an inherent feature, not a bug, of the product.

Even as AI applications spread rapidly, liability for AI systems going wrong remains an open question. But despite the rise of attention and scholarship around artificial intelligence since the “GenAI explosion” sparked by ChatGPT’s release in fall 2022, researchers have so far overlooked nondeterminism’s profound consequences for the law of AI.

This article argues that nondeterminism is the key link to successfully arguing a host of tort claims against the creators and deployers of AI products when these products cause injury. The inherent dangers of deploying a nondeterministic, LLM-based system give rise to multiple potential tort claims, especially when AI systems are deployed in contexts that aresafety-critical (where the cost of individual errors is comparatively high) oragentic (when AI agents have comparatively few checks from humans or deterministic software). Even if the average legal reader isn’t aware of nondeterminism, the average AI engineer is. By deploying a system that is known to be nondeterministic and unpredictable, developers might be accepting responsibility for the harms that emerge from a model’s behavior.

Nondeterminism is the critical hook into multiple popular tort doctrines, especially negligence and product liability (design defect) claims. In claims of negligence, developer knowledge of nondeterminism helps establish a duty and makes even nominally unexpected harms foreseeable, as they are downstream from an unpredictable, nondeterministic system. In product liability, nondeterministic systems deployed in safety-critical or agentic contexts might be a defective design for their use case. Quality assurance (QA) procedures, critical to ensuring the safety of mission-critical systems, are inherently incomplete for nondeterministic systems. Nondeterminism might even make deploying an LLM-based system in such contexts an abnormally dangerous activity, further strengthening the case for a strict liability regime.

This piece arrives at a critical moment in the development of common law around AI. AI might be eating the world, but regulation has not kept pace. Little federal legislative movement is expected, leaving soft law (industry self-regulation) as a stopgap. But the rewards for AI companies deploying quickly in an era of skyrocketing valuations will likely prove too large to ignore. Without a clear liability regime, firms see little costs to counter the lucrative benefits of capturing the AI market. Absent additional regulation, the common law as it stands today is the best tool in today’s toolkit to guard against AI harms.

The article offers a novel, mechanistic approach to assessing AI liability in high-stakes domains: beginning with technical and use case specificity. By looking at the technical features of products on the market today, we can discover ways to use existing doctrine to regulate new technologies. Inquiries into the law of AI must begin with how AI systems work. This framework could drive meaningful reforms, helping strengthen the effort of using tort law to guard against AI harms without having to rely on speculative future harms. If utilized, the arguments within this piece might incentivize AI product developers—the least cost avoiders—to only deploy agentic GenAI systems when the benefits exceed the costs.

Beatriz Botero Arcila (Institut d’Etudes Politiques Paris (Sciences Po) Sciences Po Law Ecole Droit Sciences Po) has posted “AI Liability Along the Value Chain” (Published by Mozilla Foundation) on SSRN. Here is the abstract:

Policymakers around the world are increasingly preoccupied with identifying mechanisms to better assign accountability and liability throughout the AI value chain. Particularly in the EU, discussions around civil liability and AI received significant attention after the proposal of an AI Liability Directive (AILD) in 2022. While this proposal was recently withdrawn by the European Commission, the challenges posed by AI for civil liability and harmed individuals’ ability to seek redress remain more relevant than ever amid increasing adoption of AI across sectors. 

This report thus seeks to provide more conceptual clarity to these challenges and provide recommendations on what an effective AI liability framework could look like. Though it is common to think of AI systems as a singular tool, AI systems are often developed and deployed in a value chain that involves numerous actors that participate throughout the stages of creation, fine-tuning, and implementation of these technologies, or that sell and supply key components such as pre-labeled data. 

When designing a liability system for this type of multi-party scenario, there are many questions to consider: should all parties in the value chain be held equally liable when harm occurs? Or should each actor only be held liable for the extent to which they are responsible? How easy is it to establish the contribution of each party? (Spoiler alert, it may be very hard.) Another question lawyers will be familiar with is what is the right standard — should AI actors be held liable only when they fail to take the right safety measures? Or should they be held liable regardless of whether they took safety measures, simply because by developing or deploying an AI system or model they created a risk? 

This Report discusses these questions and the complexities of assigning liability along the AI value chain, given the involvement of multiple actors in the design, development, and deployment of AI systems. The Report explores various configurations of AI value chains, the roles of different actors, and how companies allocate liability amongst them via contracts and terms. It then examines different policy choices for designing liability regimes.

Cristian Trout (Artificial Intelligence Underwriting Company) has posted “When Does Regulation by Insurance Work? The Case of Frontier AI” on SSRN. Here is the abstract:

No one doubts the utility of insurance for its ability to spread risk or streamline claims management; much debated is when and how insurance uptake can improve welfare byreducing harm, despite moral hazard. Proponents and dissenters of “regulation by insurance” have now documented a number of cases of insurers succeeding or failing to have such anet regulatory effect (in contrast with a net hazard effect). Collecting these examples together and drawing on an extensive economics literature, this Article develops a principled framework for evaluating insurance uptake’s effect in a given context. The presence of certain distortions – including judgment-proofness, competitive dynamics, and behavioral biases – createspotential for a net regulatory effect. How much of that potential gets realized then depends on the type of policyholder, type of risk, type of insurer, and the structure of the insurance market. The analysis suggests regulation by insurance can be particularly effective for catastrophic non-product accidents where market mechanisms provide insufficient discipline and psychological biases are strongest. As a demonstration, the framework is applied to the frontier AI industry, revealing significant potential for a net regulatory effect but also the need for policy intervention to realize that potential. One option is a carefully designed mandate that encourages forming a specialized insurer or mutual, focuses on catastrophic rather than routine risks, and bars pure captives.

Bryant Walker Smith (U South Carolina Joseph F. Rice Law) and Matthew Wansley (Yeshiva U Benjamin N. Cardozo Law) have posted “Regulating Robotaxis” (99 Southern California Law Review 603 (2026)) on SSRN. Here is the abstract:

In several sunbelt cities, commercial robotaxi service has arrived. The leading robotaxi company is providing over 400,000 trips per week. The industry claims that robotaxis will save lives and provide convenient and affordable mobility. Critics counter that they will increase congestion, undermine transit, and subject the public to ubiquitous surveillance. We argue that the social impact of robotaxis depends on how they are regulated. We emphasize two points missing from the debate. First, some of the benefits of robotaxis may be political rather than technological—some longstanding public policy goals may become viable in a robotaxi world. Second, letting one private company dominate the transportation system risks monopoly abuse—and regulators can act now to prevent it.

In this Article, we offer a plan to regulate robotaxis. Carefully crafted externality regulation can address pollution, congestion, wear-and-tear on infrastructure, and privacy risks while minimizing distortions in choices between travel modes. Regulators can promote competition by permitting open entry, banning lock-in contracts, and enabling one-stop access to competing networks. And they can protect riders even if competition fails by mandating that fares be transparent and rider-neutral and requiring that robotaxi companies maintain a fleet sufficient for emergencies. Policymakers should take advantage of robotaxi deployment to reimagine the transportation system—liberate land from the tyranny of parking, refocus mass transit investments on high-throughput routes, and expand mobility for people with low incomes and people with disabilities.