The use of AI to make autonomous consequential decisions about people, informed by – but often replacing decisions made by – human driven bureaucratic processes, has led to concerns about liability.