Key points are not available for this paper at this time.
The point that Artificial Intelligence (AI) will change the law is trite and obvious by now. How it will change the law, and how the law will change AI, are much harder questions to answer, however. Most of the hard questions arise when what I have elsewhere called an ‘autonomy threshold’ has been crossed.1 One could also call it an agency threshold. What this means is that when an AI machine, whether in the form of a robot (such as a drone or self-driving vehicle), or a bot or other software-implemented AI system makes choices, the legal situation changes. To an outside observer, a self-driving car unquestionably makes choices: did it avoid an accident by going left or right? What if it caused a smaller accident to avoid a bigger one – that is the well-known trolley problem traditionally used by ethics professors. Machines make all kinds of choices. An AI machine is reportedly used in Canada to decide whether the marriage of someone applying for a residence permit after marrying a Canadian is a real marriage. AI machines suggest bail and sentencing decisions to judges, who by and large follow them. And the list goes on. The law will have to apportion liability for the choices made by those machines, their owners, and their users/operators because the choices made or informed by those AI systems impact real human lives.
Daniel J. Gervais (Sat,) studied this question.