The Rise of Algorithmic Decision-Making in Warfare

Artificial intelligence is increasingly being used in the military for planning and operations as a decision support tool at multiple stages. The US’s useof Anthropic’s Claude model against Iran marks a significant moment in the history of warfare. Integrated via Palantir’s Maven Smart System, AI-supported intelligence analysis, target identification, and operational simulations enabled planners to process information faster than human capabilities. While analysts have framed this as an “AI war,” the more significant shift lies in the growing influence of algorithmic systems in shaping military decision-making architectures.

Admiral Brad Cooper, who led Operation Epic Fury, said that AI systems processed massive amounts of intelligence and surveillance data, allowing commanders to gain insights within seconds. This is part of a wider movement to shift more complex intelligence tasks to algorithmic systems, raising questions about transparency, oversight, and reliance on algorithmic assessments.

This is also observed in other conflict zones, but in different operational roles. In Gaza, Israel’s Lavender system, developed by Unit 8200, assisted in the targeting of 37,000 suspected individuals, based on reported affiliations, using AI. Structural strikes and real-time tracking were made possible through the use of additional tools like “The Gospel” and “Where’s Daddy?” These systems reduced human review into quick, seconds-long “stamp of approval” decisions, moving targeting to machine-driven validation. In Ukraine, AI tools were used to assist in drone operations and battlefield analysis by training datasets. Initial programs, like Project Maven, relied on manually labeling 150,000 images. Currently, the Brave1has enabled over 100 defense-tech firms to train combat AI on millions of annotated images from ongoing missions to improve these AI models.

The modern battlefield produces unprecedented volumes of data from interwoven sensor networks, drones, satellite imagery, and localized communications streams. This information comes at high speed and volume, which can overload the human brain. AI is being used to deal with this information overload, but there are concerns about the accuracy of AI-driven assessments and how much human oversight might be required to rely on AI. Military officials emphasize that humans have the final authority, but systematic integration poses challenges to oversight quality. The other predicament is automation bias, a psychological phenomenon in which a human operator, particularly under pressure or high stress, is likely to rely on the system’s recommendations. Therefore, striking a balance between speed and responsibility, ethical judgment, and accountability in the use of force is a key challenge.

Another area of concern pertains to legal and ethical issues. International humanitarian law is based on the principles of distinction, proportionality, and precaution. With the growing use of AI in military operations, it becomes more difficult to apply these principles, thereby making accountability and scrutiny more difficult. The International Committee of the Red Cross has warned that, when algorithmic systems provide input for analysis, targeting, or operational planning, it is hard to assign responsibility for any errors. Even with humans “in the loop,” the black-box nature of machine learning limits transparency and complicates legal review. It is not just a theoretical problem; it has been seen in practice. In the early US campaign against Iran, an AI-assisted missile struck a girls’ school near an IRGC compound, killing 120 children, likely due to a classification error. Anthropic’s CEO’s admission of limited awareness over Claude’s use in the strike highlights a broader issue. AI developers are fully aware of the risks associated with delegating autonomous functions to AI, yet they continue to promote its adoption. As AI assumes greater decision-making roles, concerns over misidentification and the possibility of AI acting against human directives are often overshadowed by narratives emphasizing its benefits.

For Pakistan, these developments are neither distant nor theoretical. In a region where crises can escalate quickly, AI-enabled decision support offers advantages but also carries risks. It improves situational awareness and accelerates analysis but compresses decision time, limits verification, and heightens the risk of miscalculation. Considering both, Pakistan is accelerating efforts to build AI capacity and strengthen its supporting infrastructure. At the policy level, this translates to a recognition that successful adoption is not just about adopting algorithms but about enhancing data governance, institutional maturity, and a skilled workforce capable of embedding AI into decision-making processes. Thus, Pakistan’s approach remains focused on leveraging AI to bolster human judgment in intelligence fusion, surveillance, logistics, and cyber defense.

There is a clear lesson from the academic literature and initial operational experience: algorithmic systems are transforming military information processing. However, as their role in decision-making grows, they also entail bias, error propagation, lack of transparency, and overreliance on machine-generated recommendations. AI, therefore, must be used as a support system, with humans retaining final decision-making responsibility. This requires investment in training, auditability, and institutional safeguards to ensure that human decision-makers are meaningfully engaged, rather than merely present in form. The future of warfare will likely be defined not by machines acting alone, but by humans making increasingly time-pressured decisions shaped by machine-generated insights. The central strategic challenge is not whether to adopt algorithmic tools, but how to ensure that their speed never outpaces sound judgment.