Attributing Liability in AI-Integrated Legal Services

When an AI tool gives a lawyer or a client wrong legal advice, who should pay? AI providers write almost all liability out of their terms of use, and under a fault-based rule they could pass off preventable errors as unavoidable hallucinations. This paper argues that they should instead be strictly liable, with a defence for careless users, and that the EU’s revised Product Liability Directive can already be read this way.