When a device controlled by artificial intelligence (AI) is involved in an accident causing bodily harm, who is at fault? The manufacturer, software developer or the user?

With so many people involved in the design and production of AI-controlled technology (known as the 'problem of many hands'), liability can be difficult to attribute, leaving victims and their families without recourse. As AI becomes increasingly entrenched in our daily lives, mishaps will become more common.

This policy brief looks at the challenges in determining who is at fault when AI goes awry and how to address the liability gap.

Key points:

• Traditional tort liability frameworks are unsuitable in addressing accidents involving artificial intelligence (AI) systems due to the unpredictability of AI outcomes and the opacity of AI systems, resulting in a liability gap.
• The European Commission has proposed changes to address the challenges, including measures to alleviate the evidentiary burden on claimants by decreasing the burden of proof and issuing disclosure orders.
• In cases involving high-risk AI systems, strict liability and no-fault liability insurance should be utilised. The presence of insurance covers, however, may encourage actors to increase their exposure to risk.