The age of artificial intelligence (AI) has arrived on the battlefield through a steady accumulation of capabilities that are reshaping how militaries see, decide, and act. The promise is undeniable: faster decision-making, sharper intelligence, and reduced risk to human soldiers.
Yet in the midst of the rush that AI is the way forward for future warfare, two issues need to be considered: maintaining meaningful human control and managing its role in intelligence and targeting, without undermining accountability.
Human control
The first challenge is meaningful human control: it means that humans remain responsible for decisions involving the use of force.
Even with rapid advances in AI, human decision-making remains essential within the OODA (Observe–Orient–Decide–Act) framework because humans uniquely interpret context and ambiguity in ways machines cannot fully replicate. While AI can accelerate observation and analysis, the “Orient” and “Decide” stages often require judgment shaped by ethics, experience, cultural awareness, and long-term intent, factors that are difficult to encode into algorithms. In this sense, AI enhances the OODA loop, but human oversight ensures decisions remain adaptable, accountable, and aligned with real-world consequences.
In practice, however, AI systems increasingly compress decision timelines to the point where human intervention risks becoming nominal. When an algorithm processes sensor data, identifies a target, and recommends action within seconds, what does “control” actually look like?
A human approving a pre-packaged recommendation is not exercising judgment in any meaningful sense; they are endorsing the output of a system they may not fully understand.
This creates a dangerous illusion of oversight. Commanders may believe they are in control while they are actually delegating critical judgments to opaque systems trained on imperfect data.
A case in point is the AI-based decision support systems (AI DSS) that are being used. AI DSS consolidate relevant information and integrate it with AI capabilities to support analysts, operators, and key stakeholders in processing large volumes of data, thereby enabling informed decision-making on matters ranging from movement of forces to targeting of enemies.
The challenge here is to ensure that outputs from rAI DSS should support rather than replacing the human commander in making decision. However, humans tend to over-trust machine outputs, especially under pressure. In high-tempo operations, the temptation to defer to AI recommendations will only grow. Ensuring meaningful control, therefore, requires more than keeping a human “in the loop.” It requires redesigning workflows so that humans retain genuine decision authority, along with the time, information, and training needed to exercise it. Otherwise, there will be a diffusion of accountability.
ISR & Targeting
The second challenge emerges from AI’s expanding role in intelligence, surveillance, and reconnaissance (ISR). AI enables the processing of vast amounts of data such as satellite imagery, signals intelligence, drone feeds, far beyond human capacity. This can enhance situational awareness, such as sending immediate alert to the ground commander when there is a significant event in the Area of Operations.
However, the use of AI in ISR introduces new vulnerabilities. AI systems can misinterpret data and inherit biases from training sets. According to a study by the Stockholm Institute Peace Research Institute, the sources of bias in military AI come from bias in society, data processing and algorithm development, and use.
The report posits that societal bias may stem from distorted media representations and the underrepresentation of certain population groups in available datasets. It further identifies bias in data processing and algorithm development as emerging from the assumptions embedded by system designers, for instance, the interpretation of a human posture with raised arms as a proxy for surrender. Additionally, the report highlights bias in use, which arises during the deployment phase of military AI systems. In such cases, systems employing positive feedback loops may internalise the preferences of individual users, leading algorithmic outputs to increasingly reflect and potentially reinforce those users’ subjective biases.
Moreover, as ISR systems become increasingly automated, there is a risk that analysts may shift from active evaluators to passive consumers of intelligence outputs. The core analytical practices, such as interrogating assumptions, corroborating sources, and situating information within its broader context, may erode if AI-generated outputs are treated as authoritative.
Research conducted by the MIT Media Lab suggests that excessive reliance on AI-driven systems may inadvertently contribute to cognitive atrophy. Complementary studies indicate a significant negative correlation between the use of AI tools and the development of critical thinking skills, with younger users demonstrating higher levels of dependence and correspondingly lower cognitive performance scores.
A major negative consequence of reduced cognitive performance, combined with AI-driven sensors and decision-support inputs, is the erosion of basic soldiering skills, the ability to make sound judgments under harsh and rapidly changing conditions. On the battlefield, mental fatigue is a reality. Combatants often operate under extreme sleep deprivation while juggling multiple tasks amid physical exhaustion, stress, and constant threat. These vulnerabilities can have tragic and irreversible consequences. A tragic example is the US strike on a Doctors Without Borders trauma facility in Kunduz City, Afghanistan, in 2016, leading to the death of 42 people. Several factors leading to the error was “fatigued from days of fighting” on the part of the Americans.
In sum, the Kunduz tragedy underscores how human limitations, not just technological flaws, shape outcomes on the ground. Even highly trained personnel, operating under fatigue and intense pressure, can misidentify targets and act on incomplete, outdated or inaccurate information. As AI systems increase the volume and speed of information, these cognitive burdens will only intensify. Without careful safeguards, reliance on AI risks amplifying, not reducing, such errors, with potentially devastating consequences.
While AI has already transformed modern warfare, a critical question remains: can militaries effectively harness its advantages without sacrificing human judgment, accountability, and restraint? These principles, though imperfectly upheld, have long been central to the ethical conduct of war and must not be sidelined.
