Policy Brief No. 218 — December 2025
Trust by Design? AI in Military Applications Branka Marijan
Key Points → Military adoption of artificial intelligence (AI) is outpacing governance, compressing decision time and heightening escalation risk. In response, “trust” must be operationalized as assured reliance, clear claims, evidence and accountability — at system, organizational and interstate levels. → “Trust-by-design” frameworks offer possible ways in which to ensure both technical and governance considerations are embedded into the life cycle of AI systems. → Confidence building measures (CBMs) between states, notably the United States and China, on emerging military technologies are necessary to ensure trust building. CBMs such as mutual declarations of principles, incident notification protocols and joint verification mechanisms are urgently needed to prevent miscalculation and escalation.
Introduction As rivalries sharpen from Washington to Beijing, a new fault line is emerging: trust in AI. Nowhere is the question more pressing than in military applications and international security discussions. Militaries want systems that soldiers will use, and commanders can rely on. Diplomats want frameworks that can prevent accidents and reassure adversaries. At stake is whether frameworks, such as “trust by design,” can bridge these needs across technology and governance. The issue of trust in AI extends beyond defence. A 2025 University of Melbourne and KPMG survey of 48,000 people in 47 countries found that more than half, some 54 percent, remain wary of trusting AI (Gillespie et al. 2025). Concerns range from accuracy and bias to contextual blind spots and the risks of misuse. Interestingly, some 70 percent of those respondents see regulation as crucial in addressing the mistrust (ibid.). A calibrated level of trust is essential: too much trust can lead to overreliance on AI, which itself carries risk, including in unexpected ways, such