Legal reviews of new weapons, means, and methods of warfare, particularly those incorporating Artificial Intelligence (AI) in the military domain, are crucial for ensuring compliance with international humanitarian law (IHL). This article argues that while such legal reviews are essential due to international legal obligations and their role as confidence-building measures, they are ultimately insufficient on their own. The unique characteristics of military AI, such as its broad application, iterative development, issues with reliability, and the significant role of the private industry, limit the effectiveness of traditional review methodologies. Moreover, legal reviews are procedural mechanisms that interpret existing law, rather than normative tools capable of establishing new standards or fostering international consensus on AI governance in armed conflict.
Essential as a Legal Requirement
Legal reviews are legally required under international law, primarily by Article 36 of Additional Protocol I to the Geneva Conventions. This mandates that states assess new weapons, means, or methods of warfare for compliance with international law before deployment. Even for states not bound by API, the due diligence aspect of Common Article 1 to the Geneva Conventions supports this obligation, ensuring states take reasonable steps to prevent IHL violations with military AI capabilities.
Essential as a Confidence-Building Measure
In the absence of an international verification regime for military AI, national legal reviews serve as the primary mechanism for states to not only ensure but also demonstrate that their military AI capabilities comply with international law. Transparency regarding the review process, even without disclosing technical details or specific outcomes, is vital for building confidence among states regarding the responsible use of AI in the military.
The Features of Military AI that Limit the Effectiveness of a Legal Review
Military AI possesses unique characteristics that hinder the effectiveness of traditional legal review methodologies. These include the broad applicability of AI across various military functions, necessitating a wider range of legal considerations; the iterative development process of AI, which makes a single pre-fielding review inadequate; challenges in foreseeing AI's effects due to its opacity, brittleness, and vulnerability to adversarial tampering; and the significant role of private industry holding critical information and expertise, which can limit state access for comprehensive reviews due to proprietary concerns.
Legal Review as a Procedural Mechanism
Despite their importance, legal reviews are fundamentally procedural mechanisms. They operate within and reflect existing law, facilitating compliance with current obligations rather than developing new ones or promoting a common international understanding of legal interpretations. Therefore, they cannot address the normative questions central to the policy debate on military AI, such as determining appropriate standards, human involvement requirements, or prohibited applications.
Concluding Thoughts
Legal reviews for military AI are both essential and insufficient. They are necessary for ensuring legal compliance and fostering confidence among states, especially given the lack of international verification. However, the inherent characteristics of military AI limit their standalone effectiveness, and their procedural nature means they cannot substitute for establishing new international normative standards. Nonetheless, these reviews generate practical experience and technical understanding critical for any informed policy discussions and for implementing future collective normative choices.
