Governments are beginning to use algorithms to influence decisions affecting citizens' rights, benefits, security and opportunities. But when an AI system becomes part of the machinery of government, who is legally accountable for its decisions?