The wars in Ukraine, Gaza and Iran are not only humanitarian crises — they are laboratories for a new kind of warfare. Across all three conflicts, artificial intelligence (AI) is reshaping how targets are identified, how quickly decisions are made, and who holds power over life and death. The result is a battlefield increasingly shaped by data, algorithms and machine-assisted judgment — and governed, for now, by no one.
Autonomous and semi-autonomous drone systems developed by firms such as Anduril Industries are increasingly used for surveillance and strike missions, leading to a rapid shift toward machine-assisted combat. In Ukraine, companies such as Palantir Technologies have provided AI-enabled platforms that fuse satellite imagery, drone feeds and battlefield data to support real-time targeting and operational planning, while Clearview AI has been used for facial recognition in identifying combatants and casualties and in uncovering Russian combatants. Anthropic’s AI model Claude was used by the Pentagon during the operation to capture Venezuelan leader Nicolás Maduro. Meanwhile, major cloud providers such as Microsoft supply the computing infrastructure and AI tools that enable large-scale data processing, logistics optimization and intelligence analysis.
The use of AI in war poses questions about human control and decision making, the role of technology in society and the relationship between human and machine.
Reducing the Decision Loop
At its core, AI in warfare functions as a system: programs developed through initiatives such as Project Maven — a US Department of Defense initiative launched in 2017 to accelerate the adoption of AI and machine learning in military intelligence — integrate hundreds of data sources and use machine learning to identify patterns, suggest targets and support decision making. In the context of war, AI can connect raw data and military action, allowing armies to operate at a speed and scale that would be impossible for humans in real time.
Using AI in war can be very helpful and legitimate within a certain framework of rules, but its broader effects may be more destabilizing. By reducing the time required to identify and strike targets and lowering risks to armed forces, AI may also inadvertently lower the threshold for the use of force. Faster decision making can come at the expense of deliberation, thus increasing the likelihood of escalation. As a result, violence and warfare may become more frequent and scalable, while human oversight is diminished.
Despite official assurances that humans remain “in the loop,” the reality is more complicated. AI systems increasingly shape the options that human operators see and thus base their decisions on, prioritizing targets and recommending actions. For example, the United States’ Project Maven has the capacity to analyze massive volumes of drone surveillance to identify individuals who are armed, correlate movement patterns with prior intelligence and then potentially prioritize an individual as a high-value target.
A joint investigation by +972 Magazine and Local Call found that the Israeli military has made extensive use of AI in its operations in Gaza. Central to this effort is a system known as “Lavender,” which analyzes behavioural patterns, social networks and other data to assign individuals a level of suspicion. In the early weeks of the conflict, the Israeli military reportedly relied heavily on the “Lavender” system, which identified around 37,000 Palestinians as suspected militants and marked their homes as potential targets. According to the report, approvals were often made in under 20 seconds, with minimal human review despite a reported 10 percent error rate.
In wars, decision makers are faced with time pressure and information overload that makes them more likely to defer too much to these AI systems. This is also known as automation bias, a tendency for humans to rely excessively on automated systems and defer to outputs produced by such technologies. As a result, human oversight may eventually become more procedural. Systems such as “Lavender” and “The Gospel,” intended to streamline military targeting, have raised serious legal and ethical concerns. Investigations by The Guardian point to a strong reliance on AI with minimal verification, permissive oversight and decisions often justified by statistical confidence rather than confirmed accuracy.
The reality is that the loop is narrowing between data collection, interpretation and the decision to act, which introduces new forms of risk. AI systems are only as reliable as the data they are trained on. For example, the strike that hit a school in Iran on February 28, 2026 and killed 175 people was misidentified as a military target due to faulty or outdated data and the rapid, automated “compressed kill chain” that left little room for meaningful human oversight. Without proper human-led verification processes, errors can now propagate more quickly. Even if humans remain in the loop, they now operate under institutional pressure to act quickly.
Moreover, many AI systems operate as “black boxes,” producing outputs that are difficult to interpret. When failures occur, such as misidentified targets or flawed recommendations, it may be impossible to trace exactly how or why the system reached conclusions due to its opacity and complex predictive power and data-driven learning. Predictive policing tools such as the COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) recidivism algorithm raised major concerns because they operated as opaque “black boxes,” making it difficult for judges or defendants to understand or challenge why someone was labelled high risk. Similarly, the Los Angeles Police Department’s (LAPD’s) predictive policing program LASER was discontinued after a 2019 review by the LAPD inspector general identified unreliable data, weak oversight and biased criteria for labelling individuals as “chronic offenders.”
AI does not necessarily eliminate uncertainty. On the contrary, it can make it more complicated to understand how decisions are made.
The use of dehumanized tools is also greater when the enemy is already dehumanized. The use of AI-assisted targeting in military attacks against infrastructure and individuals is an example of this. Human lives are turned into data, and faced with deluges of data, operators could act quickly and without much concern for the people at stake.
AI Ethics in Warfare: Who Sets the Rules?
There are profound ethical and legal questions at stake. When AI systems contribute to operational decisions, accountability becomes diffuse and existing legal frameworks are not currently equipped to address such distributed responsibility.
Defence officials and experts have warned that the existence of capabilities such as AI-enabled surveillance and control can alter the character of democratic life as the line between national security and mass domestic surveillance becomes increasingly difficult to draw, especially in wartime. Many AI applications may be legally permissible under existing frameworks yet still raise serious moral concerns. As Jake Laperruque — deputy director of the Security and Surveillance Project at the Center for Democracy and Technology — argues on Tech Policy Press, debates over defence contracts with AI companies such as OpenAI and Anthropic show that terms such as “mass surveillance” or “autonomous weapons” are often poorly defined, leaving room for broad interpretation and potential abuse.
These issues point to a deeper problem: the absence of effective governance at both the national and international levels. Corporate ethics are inherently limited because they operate within competitive markets and are subject to government pressure; one firm’s refusal can simply be bypassed by another’s willingness to comply. Moreover, it cannot and should not be the role of tech companies to be in charge of national rule setting. There should be collaboration between all actors capable of generating the tools and algorithms, those who will use them, and those with the power to make decisions.
Finally, the integration of AI into warfare is being driven not only by ongoing wars, but also by geopolitical competition, particularly between major powers such as the United States and China. This competition creates strong incentives to deploy AI rapidly, even before its risks are fully understood. The result is a potential “race to the bottom,” in which ethical safeguards are weakened in the name of strategic advantage.
While China is, publicly at least, open to AI governance in war, the United States and Israel currently seem neither inclined to put structures in place nor to address ethical and legal questions. Profit and the will to win will always be powerful forces — the real task is ensuring that they are never the overriding ones.
