— Artificial intelligence is testing long-standing product liability principles because many machine-learning systems operate in ways that even their creators cannot fully explain. When an AI-powered vehicle, medical device, industrial robot, or consumer product causes harm, investigators may be able to determine what happened without being able to explain exactly why the system made a particular decision. That challenge is becoming a central issue in litigation involving AI-enabled products.

Traditional product liability cases often focus on identifiable defects, component failures, manufacturing errors, or design choices that can be traced and explained through expert analysis. AI systems differ because their outputs are generated through complex models trained on vast amounts of data rather than explicit human-written instructions. This creates difficulties for manufacturers attempting to defend product performance and for claims professionals tasked with evaluating causation and liability.

A developing legal theory argues that the inability to explain an AI system’s decision-making process may itself constitute a design defect. Plaintiffs may contend that manufacturers should have adopted more interpretable models, stronger logging capabilities, enhanced safety controls, or human oversight mechanisms. Defendants are expected to respond that simpler, more explainable systems often sacrifice performance, accuracy, and safety benefits delivered by advanced machine-learning models. The balance between transparency and effectiveness may become a key battleground in future litigation.

Another emerging challenge involves AI systems that evolve after deployment. Unlike traditional products, some machine-learning models continue adapting to new data and operating conditions. This raises difficult questions about when a product should be evaluated for defectiveness and what obligations manufacturers have to monitor performance, identify drift, and implement corrective updates. Documentation, monitoring records, testing protocols, and update histories could become critical evidence in both claims investigations and courtroom proceedings.

For claims adjusters, these disputes are likely to involve extensive technical evidence, specialized experts, and large volumes of discoverable material. Investigations may increasingly focus on training data, algorithm testing, system logs, software updates, governance procedures, and post-deployment monitoring practices. As courts establish standards for AI-related liability, insurers and claims professionals will need to adapt traditional product liability frameworks to technologies whose decision-making processes are often difficult to interpret or explain.