The Shifting Landscape of Warfare: AI's Growing Role
The integration of artificial intelligence (AI) into military strategies is fundamentally altering the dynamics of modern conflict. From sophisticated autonomous drones to AI-driven command systems, these technologies are streamlining decision-making processes and reducing the requirement for extensive human intervention on the ground. This transformation is already evident in ongoing conflicts, with new AI applications emerging in various global hotspots. For instance, interceptor drones powered by AI have been deployed in Ukraine, while Israel has utilized AI for target generation in Gaza. More recently, the Maven Smart System (MSS) by Palantir was reportedly employed during a US-Israel conflict with Iran to identify and strike numerous targets.
During the initial phase of the US-Israel conflict with Iran on February 28, the Maven system reportedly facilitated strikes on over 1,000 targets. Among these was an Iranian primary school in Minab, an incident that tragically resulted in the deaths of more than 150 individuals, predominantly children. Four months later, the Pentagon's investigation into this incident remains inconclusive, with senators pressing for answers. Reports from Bloomberg suggest that the error stemmed from outdated satellite intelligence. As AI becomes more deeply embedded in military operations, the central question extends beyond the accuracy of these systems to their capability for moral reasoning and their ability to grapple with the ethical ramifications of their actions. This article explores whether AI can effectively assume the role of human decision-makers in warfare.
The Ascent of AI on the Battlefield: A New Era of 'Decision Advantage'
Global military spending saw a 2.9 percent increase last year, reaching $2.87 trillion. Within this substantial figure, investment in AI for military applications is experiencing rapid growth, with projections indicating an increase from $11.7 billion to nearly $19.3 billion by 2030, according to industry analyses. In April, the Trump administration proposed a $1.5 trillion defense budget for the upcoming fiscal year, aiming to cultivate a “dream military” and accelerate the transition from a traditional military-industrial complex to a technologically advanced one.
The US Pentagon aims to become an “AI-first fighting force,” focusing on compressing the OODA loop (Observe, Orient, Decide, Act). This decision-making model, developed by US Air Force Colonel John Boyd, posits that speed in decision-making confers a significant advantage. By continuously feeding new information, the OODA loop enables commanders to make decisions at the latest possible moment, prioritizing speed and adaptability over absolute certainty. Driving this emphasis on speed is Project Maven, Palantir's flagship AI intelligence platform, which, when paired with Anthropic’s Claude, aims to provide real-time battlefield intelligence.
The logic underpinning the OODA loop was already evident in the drone warfare of the 2000s and 2010s. During this period, algorithmic pattern recognition identified individuals whose behavior matched suspicious profiles, generating computationally derived kill lists. Although not explicitly termed AI at the time, the underlying principle was identical: swiftly identify patterns in data that a human analyst might miss and translate them into actionable intelligence.
How the Maven Smart System Functions
Understanding Maven's impact requires an appreciation of the operational changes it has introduced. In previous conflicts, such as the US wars in Iraq and Afghanistan, analysts relied on spreadsheets and presentations to track adversarial networks, logging names and mapping connections. Kill chains were dependent on printed dossiers reviewed by senior officials, processes that were time-consuming. In contrast, during Operation Epic Fury, the current US conflict with Iran, Maven has reportedly processed thousands of strikes in minutes. Admiral Brad Cooper, head of US Central Command, noted that timelines that once spanned hours or days have been reduced to mere seconds.
Project Maven, initially launched in 2017 as the Algorithmic Warfare Cross-Functional Team (AWCFT), was designed to leverage AI and machine learning to automate the analysis of vast amounts of drone and surveillance footage that the military collected but could not fully process. Its functions range from identifying and prioritizing targets to selecting appropriate weaponry and assessing battle damage. Google initially developed computer vision models for analyzing drone footage but withdrew from the project in 2018 following staff protests, with over 4,000 employees signing a petition asserting that “Google should not be in the business of war.”
Currently, more than 20,000 US military personnel utilize Maven across 35 military software tools and three security classification domains, according to Vice Admiral Frank Whitworth, director of the National Geospatial-Intelligence Agency. The system aggregates data from over 150 sources, including satellites, commercial radar constellations providing imagery through various conditions, drone videos, signals intelligence (SIGINT), radio emissions, social media geolocation, and field reports from deployed units. Once a target is identified, it progresses through a digital pipeline modeled on the military's F2T2EA (find, fix, track, target, engage, assess) kill chain sequence, complete with coordinates, intelligence sources, risk assessments, and legal sign-offs at each stage. An AI recommender suggests weapons and munitions based on proximity and suitability. Anthropic’s Claude, integrated via Palantir’s platform, allows operators to query the system using natural language. Anthropic was the first “frontier” AI company to deploy its models on the classified networks used by Maven. However, tensions arose in March when the Pentagon blacklisted Anthropic, designating it a “supply-chain risk,” after the company reportedly refused to relax Claude’s restrictions concerning autonomous weapons and surveillance, leading Anthropic to initiate legal action against the government.
The Erosion of Human Decision-Makers
The US Department of Defense maintains that no AI weapons will be fully autonomous, always requiring human authorization. Admiral Cooper stated in March that “Humans will always make final decisions on what to shoot and what not to shoot and when to shoot, but advanced AI tools can turn processes that used to take hours and sometimes even days into seconds.” However, some academics express concern over this very premise. They argue that seemingly seamless workflows, compressed timelines, and “pre-packaged” decisions risk undermining the conditions necessary for meaningful human judgment and moral decision-making.
Elke Schwarz, a professor of political theory at Queen Mary University London and author of Death Machines: The Ethics of Violent Technologies, told Al Jazeera, “There is an implicit tension in the mandate for moral deliberation – ethical deliberation, legal deliberation – which takes time, and which requires a different way of thinking about action.” She added, “You’re saying: we’re going to sacrifice a more rigorous deliberative process in the interest of speed and scale.”
Can Military AI Ethically Replicate Human Moral Choices?
The more profound question is whether AI can, even in principle, replicate moral reasoning. Schwarz unequivocally states that it cannot, a limitation not due to a lack of sophisticated large language models (LLMs), which are used for queries and data interpretation in natural language. Just war theory was developed due to the immense stakes involved in decisions concerning lethal force. This framework exists to compel restraint, born from a hard-won understanding of the true costs of war. Schwarz emphasizes that it was never intended to be a mere checklist exercise.
“Ethics is a social practice,” Schwarz explains. “It rests on the fact that we take each other’s vulnerability very seriously, and that we trust one another not to violate that unless circumstances dictate. A system is a computational system. It has no concept of the meaning of human life. It has no concept of mourning, of suffering.” Schwarz dismisses claims of agentic AI – systems capable of pursuing tasks semi-autonomously with limited human oversight – “displaying” fear or emotion as a marketing tactic. Instead, she asserts that “grappling with principles” and doubt, and “weighing of how many lives are at stake” cannot simply be programmed into a system.
The Imperative for Restraint
Analysts warn of a significant risk in military AI: outputs that sound plausible but are incorrect, based on incomplete, outdated, or corrupted data. This type of failure is reportedly what led to Maven’s erroneous strike on the Minab primary school in Iran. While Maven’s precise error rate remains undisclosed, estimates suggest the system may be less reliable than the Pentagon indicates. Schwarz contends that large language models should not be used for targeting, as they are prone to “hallucinations” and can contribute to civilian casualties. The Intercept, an independent investigative outlet, recently reported that the Pentagon reduced its Civilian Protection Center of Excellence staff from 40 to nine, while simultaneously relying on AI tools, including one built on the Maven Smart System, to expedite civilian harm assessments. Research from the Max Planck Institute for Human Development noted that delegating tasks to AI creates a moral distance, making individuals more willing to engage in behaviors they might otherwise avoid, as AI can lend a veneer of legitimacy, with existing safeguards proving insufficient.
When informed three years ago that LLMs would be used in military targeting by 2026, Schwarz remarked, “I would have said that is lunacy. And I will still say that today.” She concluded, “To pretend that an AI system can be a moral decision-maker constitutes, for me, an abdication of this uniquely human task – to weigh, understand the weight, feel the weight of a morally difficult decision for which one might bear the burden of responsibility.”
Can AI Enforce Ethical Conduct on Militaries?
Conversely, the question arises whether AI could practically constrain less ethical militaries by imposing top-down restraint. In 2024, the Israeli-Palestinian publication +972 Magazine and Hebrew-language media outlet Local Call reported that the Israeli army was isolating and identifying thousands of Palestinians as potential bombing targets using an AI-assisted targeting system called Lavender, which had an estimated error rate of about 10 percent. Human rights experts and legal scholars warned that this practice risked violating international humanitarian law, with human oversight in some cases reduced to mere seconds before a strike was authorized.
Schwarz remains equally skeptical about AI's ability to constrain militaries. “If a military has a system that scales up and speeds up the violence they want to enact, they’re not going to abdicate to a system that says ‘no, you shouldn’t do that,’” she argued. “There’s already a problem under way that needs to be addressed beforehand. An AI system will not act as a moral decision-maker over an unethical human.”
Source: Original Article
