Report

Breaking the deadlock on AI governance

This report from UK think tank Chatham House explores how a crisis could lead to global coordination.

International AI governance is at risk of failure. Rapid geopolitical change, institutional weakness and asymmetries between the public and private sectors appear to make cooperation on AI near impossible. Proponents of inclusive, effective and global AI governance must now confront a difficult truth: that rapid progress towards global AI governance may only become politically feasible in the event of a crisis, when the costs of inaction become too great.

In this research paper, we do not claim that a crisis is desirable, but that the response to an AI crisis could lead to rapid, binding and international governance change if those involved are well prepared. Focusing investment in the right areas now will increase the chances of a productive and robust response if a crisis does come.

The research paper explains the reasons for the current deadlock and recommends actions to governments, companies and international institutions on how to respond. The paper then analyses three previous crises in other sectors – the 2007–08 financial crisis; the WannaCry ransomware attacks of 2017; and the COVID-19 pandemic – for examples of best practice and cautionary tales. The case studies indicate that crisis-driven governance works best when it brings technical expertise to the fore, and is based on pre-existing institutions and monitoring infrastructure. Attempts to build these elements from scratch post-crisis are likely to fail.