Challenges and Opportunities for AI in Military Systems

This article identifies psychological, cultural, and organizational factors that drive optimism about emerging military technologies. Psychological influences include bounded rationality, cognitive biases (like the planning fallacy and confirmation bias), and motivated reasoning. Culturally, the US military’s “can-do” ethos and historical narratives about technology’s role in victory reinforce belief in technological solutions. Organizationally, interservice competition and strategic misrepresentation by program managers seeking resources amplify optimistic projections. All of these factors combine to contribute to technological optimism in US military acquisitions. Consistent with Amara’s Law, short-term impacts are overestimated, while long-term effects are underestimated. Similar patterns often emerge as defense technology entrepreneurs consider and present “new” technologies such as AI despite mixed results. Two policy imperatives are discussed: realistic assessment through independent reviews and phased investment, and decentralized experimentation enabling rapid local adaptation alongside traditional top-down innovation.

Expectations of the performance of military technologies are marked by hopes that one’s own systems perform well while those of adversaries perform poorly, and fears of the inverse. These expectations shape states’ preparation for war and their conduct in war. But expectations frequently misalign with performance, such that the battlefield debut of novel or upgraded weapons technologies offers an opportunity for reassessment. In this article, I argue that the initial use of such weapons commonly drives a discursive process of normalization, wherein systems previously considered revolutionary or archaic are incorporated into existing modes of warfighting and accepted as normal components of those practices. I analyze the debut of several Russian long-range precision-strike weapons in the Russo-Ukrainian War, tracing the reassessment and normalization of hypersonic missiles, theater ballistic missiles, and glide bombs. This analysis shows that analysts would do well to moderate their expectations when forecasting the implications of weapons technologies.

The digital transformation of nuclear forces made modern nuclear forces more effective but potentially introduced strategic cyber vulnerabilities. Despite warnings about the cyber threats to nuclear stability, our understandings of when and why cyber operations create nuclear instability are rife with contradictory suppositions. Does entanglement create interdependence that stabilizes crisis dyads, or destabilizing pathways to inadvertent nuclear war? Do uncertainties about cyber vulnerabilities within nuclear command, control, and communications lead to a security dilemma that incentivizes preemptive nuclear use? Or does the uncertainty about how cyber operations create effects and vulnerabilities create incentives for restraint? This paper argues that current literature overlooks a foundational element of cyber and strategic stability: how the structure of networks determines the feasibility and effectiveness of cyber operations. By shifting the focus toward the intersection of network architecture and nuclear use, this piece argues that highly centralized information-processing or command nodes, which increase a network’s efficiency, can create incentives for deliberate nuclear escalation. Second, entanglement and network complexity increase the potential for inadvertent escalation or accidental nuclear use. Third, cyberattacks that exploit trust in data to degrade decision-making are the most dangerous for escalation risk.