Great Unclear Expectations: A Study on the Defectiveness of Autonomous Systems

Once autonomous systems become more widespread, product liability is expected to be called for help to compensate damages. The defectiveness of autonomous systems as a condition of product liability is difficult to assess. This difficulty becomes more evident in the case of design defects because the consumer expectations test adopted by the EU Product Liability Directive has been criticized for its vagueness, but the new Proposal for Product Liability Directive (COM(2022) 495 final) preserves it. This paper defends that the consumer expectations test is still fit for purpose and able to accommodate challenges posed by the deployment of autonomous systems. The paper first attempts to bring together the cases where there are clear safety expectations regarding autonomous systems, so the defectiveness assessment is relatively easy. In case of unclear safety expectations, it is difficult to measure and decide whether an autonomous system is defective or not. Hence, the paper’s second attempt is to make contribution to the concretization of the test in cases where it is difficult to know what to expect from autonomous systems. The public is entitled to expect the producers of autonomous systems to eliminate the harmful consequences of such products. Since the distinguishing feature of such products is their autonomy, producers are expected to consider the safety implications of the autonomy and eliminate the harmful consequences hereof. From the liability perspective, there are two main implications of autonomy that will be explored below: human–machine interaction and testing of autonomous systems. Accordingly, the difficulties arising from these two main implications should shape the scope and extent of the public’s expectations and hence the producers’ duties.