Alleged US attack on school in Iran: Palantir system in focus
Following the fatal airstrike on an Iranian school, Palantir's Maven is facing criticism.
(Image: Spyro the Dragon / Shutterstock.com)
Following the fatal US airstrike on a primary school in Minab, Iran, which killed over 160 people and is believed to have occurred on February 28 as part of a US offensive, Palantir's “Maven Smart System” is facing criticism. An analysis points to outdated data and the so-called kill chain, at the end of which the attack occurred.
Contrary to initial widespread reports, a chatbot like Claude apparently did not play a decisive role in target selection, according to current findings. The bombing, based on available analyses, suggests less an isolated technical failure and more decisions made along the entire process chain – from data maintenance to the end of the kill chain.
There are increasing indications that target acquisition was based on “Maven Smart Systems,” a military platform developed by Palantir for analyzing intelligence data. Maven links satellite imagery, sensor data, and intelligence information to identify potential targets and present them for attack approval in a highly accelerated process.
Videos by heise
According to media reports, the hit building was still listed as a military object in a US database, even though it had been used as a primary school for years. Apparently, this information was not updated – with fatal consequences. Maven is said to have adopted this incorrect classification, and it was fed into the automated decision-making process without verification.
Palantir's software is designed to massively accelerate the so-called kill chain. In military exercises, thousands of target decisions could be made in a short time. Critics, however, warn that this speed can come at the expense of human control and careful review.
The analysis, published on the Substack platform and now also in the Guardian, suggests that a system like Maven could have played a central role in this specific case. According to this, it was not an AI language model that was responsible, but the combination of outdated data and a highly accelerated, software-supported targeting process. The error, according to this account, originated where a military database entry was not updated for years and was subsequently processed by a system like Maven without sufficient human review.
Pentagon plans further AI expansion
Meanwhile, the US Department of Defense plans to further expand Maven and permanently anchor it in military strategy. According to an internal memo, the system is to be established as an official “Program of Record,” which would secure long-term funding and deployment across all branches of the armed forces. The decision could be made before the end of the current fiscal year in September.
Maven was originally launched in 2017 for the automated evaluation of reconnaissance and drone images. In recent years, Palantir has received multi-billion dollar contracts for the further development of the system. At the same time, the military, and industry are pushing ahead with the integration of AI into military operations. Pentagon officials emphasize that future conflicts will be largely data-driven and supported by AI.
(mack)