Controlling Military Artificial Intelligence Harnessing Rules of Engagement and Military Directives

In November 2024, the DILEMA and ELSA Lab Defence Projects convened a multidisciplinary team of legal, military, and technical experts at the Asser Institute in The Hague to explore how Rules of Engagement (ROE), alongside other military directives, could be harnessed control to the use of military artificial intelligence (AI). The central challenge: Can military protocols effectively address the unpredictable and rapidly evolving nature of AI in modern warfare? The workshop focused on two primary questions. First, it considered whether traditional numbered ROE could serve as an effective instrument for issuing AI-specific instructions. Second, it explored how authorities responsible for issuing these military directives would interpret and translate complex technical and operational elements into clear, actionable guidelines for the armed forces.

The outcome? While ROE remain foundational to military operations, experts agreed that AI’s inherent complexity and unpredictability may require a new approach - one that calls for flexible context-specific directives capable of addressing the unique challenges posed by this technology. This policy brief underscores the importance of developing such AI-specific directives – crafted through the strategic selection and drafting of various military frameworks – that ensure effective control and compliance with international law. A fictional use case enabled experts to engage directly with the real-world complexities of drafting such guidelines, offering practical examples of how existing frameworks can be adapted to regulate AI systems.

Providing extensive, detailed and actionable recommendations, this brief serves as a critical resource for drafters, offering guidance on how to approach, analyse, and formalise military directives for AI. It advocates for the prioritisation of AI-specific directives that not only uphold legal standards but also improve operational control and efficiency. These findings offer a comprehensive framework for navigating the evolving landscape of military AI with foresight, precision, and responsibility.