Leading in Artificial Intelligence through Confidence Building Measures

Abstract

The role of artificial intelligence (AI) in military use has been the subject of intense debates in the national security community in recent years— not only the potential for AI to reshape capabilities, but also the potential for unintentional conflict and escalation. For many analysts, fear that military applications of AI would lead to increased risk of accidents and inadvertent escalation looms large, regardless of the potential benefits. Those who are concerned can cite a plethora of potential ways things can go awry with algorithms: brittleness, biased or poisoned training data, hacks by adversaries, or just increased speed of decision-making leading to fear-based escalation. Yet, given its importance for the future of military power, it is imperative that the United States moves forward with responsible speed in designing, integrating, and deploying relevant military applications of AI. How should the United States simultaneously pursue AI swiftly while reducing the risk of unintentional conflict or escalation in the United States or elsewhere? The answer may lie in US leadership to promote responsible norms and standards of behavior for AI as part of a series of confidence-building measures (CBMs) tailored to reduce the likelihood of these scenarios.