Regulatory Markets for AI Safety

TL;DR

This paper introduces a novel concept of global regulatory markets for AI safety, illustrating how such markets could mitigate risks like adversarial attacks on AI systems used in commercial drones.

Contribution

It proposes a new regulatory framework based on global markets and demonstrates its practical application in enhancing AI safety against adversarial threats.

Findings

Conceptual model of global regulatory markets for AI safety

Analysis of costs and benefits of the proposed approach

Practical example involving adversarial attacks on drones

Abstract

We propose a new model for regulation to achieve AI safety: global regulatory markets. We first sketch the model in general terms and provide an overview of the costs and benefits of this approach. We then demonstrate how the model might work in practice: responding to the risk of adversarial attacks on AI models employed in commercial drones.