Washington (United States) (AFP) – A Pentagon artificial intelligence program, Project Maven, is at the heart of the military campaign against Iran. A unique role which illustrates a paradigm shift for modern warfare, propelled by the innovations of Silicon Valley, itself riddled with profound ethical questions.
Published on: 05/04/2026 - 08:21
. What is this ?
Project Maven is the US military's flagship artificial intelligence program, launched in 2017 as an experiment to help military analysts process the massive stream of images sent by drones.
The operators were overwhelmed, forced to work frame by frame to identify pieces of information that might only appear for a fraction of a second. Maven was designed to help them find the needle in the haystack.
Eight years later, the program has greatly expanded. It has become an AI-assisted targeting and battlefield management system, which has increased the speed of execution of the “kill chain”, this process which goes from detection to destruction (chain of engagement, in NATO jargon).
. How does it work?
Maven combines the functions of war air control and cockpit.
Expert Aalok Mehta describes the system as "essentially an overlay" merging sensor data, satellite imagery and intelligence on friendly and enemy forces.
Concretely, explains the director of the Wadhwani AI Center, based in Washington, Maven analyzes satellite feeds at high speed to detect movements or identify targets, while “drawing up an instantaneous picture of the operational theater” to determine the best strike plan.
“As if by magic,” Maven turns a detected threat into a targeting process, evaluating available solutions and presenting the command with a range of options, a Pentagon official described during a recent online demonstration.
The emergence of generative AI over the past three years has constituted a new leap forward by making it possible to interact with the system in natural language, thus democratizing the use of this technology beyond army technicians.
This capability is provided by Anthropic's Claude model, but perhaps not for long, since the American start-up was sanctioned by the Pentagon after publicly rejecting the use of its AI tools for fully automated strikes or for surveillance of American citizens.
. Why did Google say no?
Ethical questions about AI were already sensitive in the early years of the program, when Google was the original provider.
In 2018, more than 3,000 employees of the Mountain View giant signed an open letter to denounce a contract that crossed a red line. Several engineers had resigned.
Google then refused to renew the contract, then published an AI charter which excluded any participation in weapons systems.
This episode highlighted a fault line within Silicon Valley, between engineers who consider autonomous targeting to be an insurmountable ethical limit and senior military officials who consider it essential.
However, Google has just removed its restrictions and announced that it is becoming more involved in military contracts.
Google, as well as OpenAI, Anthropic's number 1 rival, and Elon Musk's xAI company are in the running to replace Claude in Maven, the Pentagon said.
. What role does Palantir play?
In 2024, Palantir, a company founded in part with funds linked to the CIA and built from the outset around intelligence, took the place left vacant by Google.
The company has become the main provider of the Maven project, and its AI technology constitutes the operational backbone of the program.
For its CEO Alex Karp, the world is now divided in two, between those who have this technology and those who do not.
It is essential, according to him, that the West manages to master capabilities that the rest of the world does not possess.
A system capable of compressing several hours of "kill chain" into a few seconds renders an adversary obsolete, he added.
. What results so far?
The Pentagon and Palantir declined to comment on Maven's performance in the war with Iran.
The brisk pace of U.S. strikes shows that Maven likely accelerated the process of targeting and firing.
During the first 24 hours of Operation Epic Fury, US forces struck more than 1,000 targets. The deadly strike that hit a school in a former military building that day was one of them, according to reports from several media outlets. A Pentagon investigation is underway.
© 2026 AFP
The content you are trying to access does not exist or is no longer available.




