The Chair AI Regulation shares a recently published article, by its research fellow, Dr. Theodoros Karathanasis, focusing on the critical intersection of Artificial Intelligence (AI) in military applications and International Humanitarian Law (IHL).
This article, titled “AI-Enabled LAWS: From the Target Recognition Principle to Adaptive Legal Reviews”, addresses the emergent challenges posed by the development and implementation of military AI, particularly concerning the protection of civilians under IHL. It highlights the inadequacy of current testing methods to address the risks associated with AI-enabled Lethal Autonomous Weapons Systems (AI-LAWS), which can undermine the reliability and predictability of these systems.
The article argues that the principle of distinction under the Law of Armed Conflict (LoAC), requiring military attacks to be directed only at military objectives, faces new complexities with AI-LAWS due to their “black box” decision-making, potential for unexpected behaviours, and degradation of accuracy over time. It expresses concerns about the quality of training data for AI-LAWS in distinguishing military from civilian targets.
Recognising the limitations of existing regulatory frameworks and the UN Convention on Certain Conventional Weapons (CCW) framework, the article advocates for the development of adaptive legal reviews that can be utilized throughout the entire lifecycle of AI-LAWS to assess the risk of target misclassification. These reviews should consider the system’s ability to comply with the principles of distinction and proportionality, its reliability, understandability, and predictability.
Cite: Karathanasis, Theodoros, AI-Enabled LAWS: From the Target Recognition Principle to Adaptive Legal Reviews (January 22, 2025). Available at SSRN: https://ssrn.com/abstract=5107346 or http://dx.doi.org/10.2139/ssrn.5107346
These statements are attributable only to the author, and their publication here does not necessarily reflect the view of the other members of the AI-Regulation Chair or any partner organizations.
This work has been partially supported by MIAI @ Grenoble Alpes, (ANR-19-P3IA-0003) and by the Interdisciplinary Project on Privacy (IPoP) commissioned by the Cybersecurity PEPR (ANR 22-PECY-0002 IPOP).
