Abstract:
Criminality and punishment have always been applied to human beings. However, the technological field of artificial intelligence ('AI') is becoming impressively sophisticated. Machines that utilise AI ('AI entities') may soon be able to commit actions which, if committed by humans, would be considered criminal. This paper poses a hypothetical fact scenario to explore whether, and how, existing criminal law should respond to such AI entities. This paper concludes that existing criminal laws are ultimately a bad fit for AI. First, regulating AI entities becomes complicated by the conceptual difficulties in defining AI. Secondly, existing party liability mechanisms, such as corporate liability, are unsuitable for non-humans. Thirdly, criminal liability has always assumed that the offender is human, meaning that AI entities cannot satisfy the mens rea element of criminality. Finally, the purposes of sentencing are so deeply rooted in society that its application to non-humans would be inappropriate. AI entities ultimately show that criminal law and social expectations are inextricably linked. This paper accordingly raises two talking points: the role of criminal law going forward, and whether AI entities will ever be accepted into the wider society.
