Strategic Stability Threats from Military AI Attacks

• •Adversarial AI attacks threaten military decision-making and strategic stability.
• •The UK is prioritizing governance and international cooperation to mitigate these risks.
• •Ambiguity in AI system failures complicates attribution and increases miscalculation risks.

The growing integration of artificial intelligence (AI) in military operations poses significant risks to strategic stability, particularly through adversarial attacks. RAND's studies highlight that these attacks could undermine the credibility of national defense, tip the offense-defense balance, and create ambiguity in distinguishing between normal malfunctions and deliberate attacks. The UK Foreign, Commonwealth and Development Office commissioned this research to address these risks, emphasizing the need for improved governance and international cooperation. Five scenarios were analyzed, revealing that adversarial AI could empower non-state actors and complicate attribution, increasing the risk of miscalculation. As AI systems become more integral to military decision-making, the consequences of adversarial interference could escalate dramatically. Recommendations include embedding adversarial AI risks in broader governance discussions and enhancing information-sharing among allies.