Companies operating AI systems are liable for their erroneous outputs just as they are for employee mistakes and cannot shield themselves through the technical nature of the system.