The governance of artificial intelligence (AI) in military contexts has become a focal point for international legal and policy discussions. At the heart of these debates lies the role of legal reviews—a mechanism designed to ensure that new technologies, particularly those with autonomous or semi-autonomous capabilities, comply with international humanitarian law (IHL). While such reviews are widely regarded as essential, recent scholarly analysis suggests they fall short in addressing the complex and evolving nature of military AI. This article explores the significance, limitations, and broader implications of legal reviews in shaping responsible AI use in warfare.

A Legal Requirement for Compliance

International law provides a clear foundation for the obligation to conduct legal reviews of new weapons and methods of warfare. Article 36 of Additional Protocol I (API) to the Geneva Conventions is the most prominent example, requiring states to assess whether the employment of a new weapon or method would be prohibited by international law in any or all circumstances. This provision ensures that military capabilities are evaluated before they are deployed, helping prevent violations of IHL.

However, Article 36 applies only to states that are parties to API. Many militarily active nations—such as the United States and Israel—are not bound by this treaty but still conduct legal reviews under their own national policies. This highlights a broader trend: even in the absence of formal international obligations, states have recognized the importance of legal due diligence in managing AI-enabled military capabilities.

Beyond Article 36, Common Article 1 of the Geneva Conventions also supports the requirement for legal reviews. While less specific than API, Common Article 1 obliges states to respect and ensure respect for IHL. This means that even if a state is not bound by API, it still has a duty to assess whether its military AI capabilities can be used in compliance with international law.

A Confidence-Building Tool in the Absence of Verification

In the absence of an internationally agreed verification regime for military AI—similar to those governing nuclear or chemical weapons—legal reviews serve as a confidence-building measure. They allow states to not only ensure but also demonstrate that their capabilities can be used lawfully.

This function relies on transparency, though it does not require full disclosure of the outcomes or technical details of legal reviews. Instead, some level of openness about how and when reviews are conducted helps build trust among nations. For instance, sharing information on the standards applied during a review or the stages at which assessments occur can reinforce the legitimacy of these processes.

Recent multilateral discussions have increasingly recognized this role, framing legal reviews as central to responsible AI behavior in warfare. These conversations extend beyond the formal scope of Article 36, suggesting that legal reviews are part of a broader normative framework for governing military AI.

The Unique Challenges of Military AI

Despite their importance, legal reviews face significant limitations when applied to military AI capabilities, which differ fundamentally from traditional weapons and software-based systems. Four key features of AI in warfare make standard review methodologies less effective:

1. Breadth of Application

AI is not a weapon but a set of techniques applicable across a wide range of military functions, including targeting, command and control, intelligence, surveillance, and reconnaissance (ISR). This breadth means that legal reviews must consider a broader array of international laws beyond IHL, such as international human rights law or the law on the use of force, depending on how the capability is deployed.

Unlike traditional weapons, which are typically assessed for compliance with specific IHL rules, AI systems may interact with multiple legal regimes. This complexity means that a review must focus not just on whether a system is lawful in design but also on the conditions under which it can be used lawfully.

2. Iterative Development

Modern AI development is highly iterative, involving continuous training and refinement throughout a capability’s lifecycle. Unlike traditional weapons, which are developed to a final specification before deployment, military AI systems evolve over time, often through updates or retraining. This makes it difficult to conduct a single legal review that covers all possible uses of the system.

This raises questions about when modifications to an AI system would require a new review. As capabilities become more dynamic and adaptable, the traditional model of conducting a review at a fixed point in development may no longer be sufficient.

3. Reliability Concerns

AI systems can exhibit behaviors that are difficult to predict or control. Unlike conventional weapons, which can be tested for accuracy and reliability using well-established methods, AI capabilities often operate in ways that are opaque, brittle, or vulnerable to adversarial manipulation. These characteristics make it challenging to foresee the effects of a system during its normal operation.

For example, an AI model trained on manipulated data may misclassify targets in ways that would not be detected through standard testing procedures. This unpredictability undermines the effectiveness of legal reviews as tools for ensuring compliance with IHL.

4. Industry’s Role

Much of the technical expertise and information required to conduct a meaningful legal review is held by private actors, such as AI developers or service providers. While states retain the legal obligation to perform these reviews, they may lack the necessary resources, access to proprietary data, or technical knowledge to do so effectively.

This creates a gap between legal obligations and practical implementation, limiting the ability of states to fully assess the compliance risks associated with military AI capabilities.

Legal Reviews as Procedural Tools, Not Normative Frameworks

Ultimately, legal reviews are procedural mechanisms that operate within existing international law. They help ensure compliance with current obligations but do not create new legal standards or promote a shared understanding of how IHL applies to emerging technologies.

Many of the most pressing questions in military AI governance—such as what standards should govern these systems, how much human involvement is required, and which applications are off-limits—are normative in nature. These issues cannot be resolved through national legal reviews alone; they require broader international dialogue and consensus-building.

Conclusion

Legal reviews of military AI capabilities are essential for ensuring compliance with international law and building trust among states in the absence of a global verification regime. However, their limitations—stemming from the unique characteristics of AI systems and the procedural nature of these reviews—mean that they remain insufficient as standalone governance tools.

As nations continue to develop and deploy AI-enabled military technologies, it is crucial to recognize that legal reviews are just one part of a larger effort to ensure responsible use. Future discussions must focus on developing normative frameworks, enhancing transparency, and fostering international cooperation to address the complex challenges posed by military AI.

Readers should watch for ongoing developments in multilateral policy discussions, as well as emerging research into how legal reviews can be adapted to better reflect the realities of modern AI systems in warfare. The intersection of law, technology, and ethics will remain a central concern in shaping the future of responsible AI use in armed conflict.

---

Original Source

This article is based on publicly available reporting. For the complete original story, visit the publisher’s article.