Service Provider Liability for Damages Caused by Generative AI Hallucinations

Abstract

The concept of damages caused by generative AI hallucinations refers to situations in which the AI generates false information, thereby giving rise to fortuitous liability. As the controllers of the AI, service providers are the appropriate parties to bear such liability. Once their status as liable parties is established, it becomes necessary to further clarify the specific rules for determining liability: in principle, fault-based liability should apply, with the state of the art serving as the standard for assessing fault; regarding the proof of causation, presumptive rules should be adopted to allocate the burden of proof reasonably. The aforementioned liability framework strikes a balance between the dual values of technological innovation and the protection of individual rights.