The conflict involving Iran has seen artificial intelligence so deeply embedded in military operations that commentators often describe it as an “AI war”: human personnel and human decision‑making have become the main bottlenecks on the battlefield. It is not only the number of soldiers but the capacity of the human brain that cannot keep pace with the incoming flow of sensor and communications data, and AI is being used to fill that gap.
How the asymmetry emerged
The architecture in use today became necessary in the early 2010s, when an asymmetry emerged between the volume of incoming intelligence and the processing capacity of human analysts. Data streams from signals intelligence (SIGINT), satellite and aerial—especially drone—imagery (GEOINT, IMINT), and open‑source intelligence (OSINT) expanded dramatically, while traditional human‑based intelligence changed little in quantity. Analysts were overwhelmed by information and AI was required to bring order to the raw data and turn it into actionable intelligence.
The kill chain and AI’s role
Modern strike processes follow what the military calls the “kill chain”: detection, collection, tracking, target designation, attack and post‑strike assessment. Technological advances have put AI into every stage of this chain, and the proportion of responsibilities delegated to machines continues to grow. The operational aim remains simple: strike as many targets as quickly and accurately as possible.
Speed and scale: Scarlet Dragon and Epic Fury
At the 2024 US exercise Scarlet Dragon, organizers ran a simulation using data from the 2003 Iraq war: the work of a previous 2,000‑person analyst group was replaced by a 20‑member staff plus AI, who designated 1,000 targets per hour and achieved an average decision time of 3.6 seconds through the chain. In the real‑world Iranians‑focused operation Epic Fury, the United States carried out more than 1,000 precision strikes in the first 24 hours, and after one week the number of destroyed targets approached 5,000. These figures contrast sharply with earlier conflicts when target identification relied on much larger analyst teams.
Risks and unintended consequences
Alongside unprecedented speed and volume, the systems in use have also produced serious errors: one of the strikes carried out by the network was the US attack on an Iranian Minabi girls’ school, which resulted in 182 civilian deaths. That incident highlights how, despite technological effectiveness, mistakes and civilian harm remain a grave risk.
Conclusion
Deep AI integration is reshaping the paradigm of strike operations: decision times and the number of engagement opportunities have increased substantially, and the demand for human analysts has fallen. At the same time, AI‑accelerated operations raise ethical and operational challenges, notably the risk of civilian casualties and erroneous targeting.
