The US Defence Department’s decision to deploy AI systems on classified military networks marks a significant shift in how modern warfare is conducted. Understanding how these systems change targeting logic and what that means for global arms competition is now an urgent strategic question.
In May 2026, the US Defence Department signed an agreement with seven AI companies to deploy their systems on classified military networks. This development is significant because it came after the United States had already used advanced decision-support systems in the recent war against Iran.
On March 11, 2026, Adm. Brad Cooper, commander of US Central Command, said, “these systems help us sift through vast amounts of data in seconds so our leaders can cut through the noise and make smarter decisions faster than the enemy can react”. According to the White House, the United States struck 13,000 targets during the first 38 days of war. This included 2,000 command-and-control centres, 1,500 air-defence targets and 1,450 industrial bases. These figures are important in understanding how these decision support systems can change the tempo, volume and logic of targeting in modern war.
AI Systems, Targeting & Weapons
There are many discussions focused on the inherent risks of these systems. For instance, their black-box nature, the possibility of data errors and automation bias, and the need for human oversight. These concerns are important and valid. However, there is even less discussion about how these systems might accelerate the broader global trends that are already making the world more unstable and are intensifying the arms competition. Global military spending has already reached 2.88 trillion dollars in 2025. This has increased by 41 per cent in the past decade. Moreover, the nine nuclear states spent 100 billion dollars on nuclear weapons in 2024; according to SIPRI, there are 12,241 nuclear weapons in the world. The real danger is not only what these systems can do on the battlefield, but how they might change what states believe they need in order to survive, compete, and win future wars.
First, states are increasing the number of weapons that may be required for AI-based targeting. These systems can process large amounts of data on military targets faster than human staff. For this reason, the capacity to identify, classify, and act on targets will increase. For example, in the US-Iran war, they reportedly hit more targets in the first four days of the campaign than they struck against ISIS in the six months. It was further reported that the US Maven Smart System brought together radar signals, satellite imagery, drone imagery, electronic communications, and other intelligence feeds into a common operating picture. As a result, it could help identify targets, generate courses of action, assess the effectiveness of strikes, and produce new target lists after previous strikes in no time.
This does not automatically increase the need for many offensive vectors. However, it changes military planning. If more targets can be identified, prioritised, and re-targeted during a campaign, states will face pressure to maintain missiles, drones, cyber tools and other platforms at a higher rate than they would without AI systems. A target list alone does not produce operational success; to achieve battlefield effects, states need weapons and platforms capable of engaging those targets. The implications of this is a global trend towards larger military spending on weapons and platforms to enhance operational success.
Attrition Infrastructure
On the other hand, the second effect of increased use of AI systems would be attrition. When Decision Support Systems (DSS) make it easier to analyse the persistent ISR feeds in real time, the chances of battlefield losses are also likely to increase. Therefore, the states will need not only weapons for the first wave of attack but also those capable of sustaining counterattacks. After the Iran war, the US Navy had already asked for a 1,200% increase in the production of the Tomahawk missiles. This stock was depleted during the US-Iran War. There is no evidence that DSS caused higher demand; however, it indicates a broader problem of military attrition. It shows that high-tempo precision warfare can consume weapons faster, and DSS-enabled targeting could intensify this pattern by generating and frequently updating target lists. Over time, this can intensify the arms race by pushing states to expand stockpiles and strengthen production capacity to prepare for a faster and more targeted heavy war.
The same targeting logic becomes more dangerous when it moves closer to strategic infrastructure. These systems can increase competition in the nuclear domain by making states worried about the survivability of their strategic assets. When states know that DSS support systems can help identify and target strategic assets or supporting infrastructure, this creates a planning dilemma: states cannot decide how many nuclear weapons, mobile launchers, decoys, communication nodes, garden sites, and TELs are enough. States will develop not only in response to the number of nuclear weapons of other states, but also to the effectiveness of those states’ DSS. This pressure will be felt most sharply by states that rely on smaller arsenals. As a result, this may deepen ongoing nuclear modernisation, as states will seek to maintain survivable forces and intensify competition in the nuclear domain.
This makes arms control even harder to maintain because strategic stability will not only depend on the number of nuclear weapons, but also on ISR networks, data fusion and algorithms. These systems will create an environment in which everyone feels unsafe. Therefore, states must take steps to ensure that these systems do not become an engine of arms racing and nuclear buildup. If states agree that DSS will not be used to target strategic infrastructure, they might feel less pressure to accelerate spending for fear that their critical assets are becoming more vulnerable. However, it will not be easy to verify whether states are not actually using these systems in targeting. To solve this problem, restraint should not only focus on algorithms themselves but also on categories, strategic infrastructure, and crisis behaviour.
More than any arms control measure, this realisation is important: DSS might provide immediate operational benefits, but whether they will translate into long-term strategic success remains a question. A world where no one feels safe cannot be stable. It is not in the national interest of any state to create an environment of continuous arms racing and nuclear buildup.
Nimra Javed is a Research Officer at the Center for International Strategic Studies AJK, working on Emerging Technologies. She holds an MPhil Degree in Strategic Studies from National Defence University, Islamabad.
