Response to the UK White Paper: A pro-innovation approach to AI regulation

Abstract

We propose a risk-based regulatory system for AI based on requiring those who develop and operate AI systems to explain and justify the steps they have taken to mitigate specified risks, and to operate their systems in accordance with those explanations.

Liability to compensate for losses caused by the AI has three bases:

* Failure to operate in accordance with the risk mitigation explanation
* Failure to perform to the same standard as a human counterpart
* Failure to take reasonable care in the development and operation of the AI, with a reversed burden of proof.

The regulatory scheme also provides for the possibility of regulatory sanctions in appropriate, high risk, cases.

Keywords: AI, Regulation, Liability, Risk

JEL Classification: K13, K2, K23, K29, K39, O38