News Critiques

What the Guardian Didn’t Explain About Amazon’s Facial Recognition Tool

The Guardian put 3.5 million photos and Amazon's facial recognition tool at the center of its latest exposé, but it never told readers what that tool can and can't actually do.

By November 18, 2026

Key Takeaways:

  • The Guardian reported on November 16 that the Israel Defense Forces uploaded 3.5 million photos of Palestinians to Amazon’s cloud but never explained that Amazon’s Rekognition tool gives a similarity score, not a match.
  • The Guardian’s report on the Israel Defense Forces’ alleged use of Amazon’s Rekognition does not say whether the military treated the tool’s results as leads that needed verification or as grounds for detention.
  • The Guardian cited a New York Times report that facial recognition in Gaza at times wrongly flagged civilians as “militants,” but that report covered a program from Tel Aviv-based Corsight, not Amazon’s Rekognition.

The Guardian published a piece on November 16 detailing the Israel Defense Forces’ (IDF) alleged use of Amazon’s infrastructure, reporting that the military uploaded 3.5 million photos of Palestinians to the retail and tech giant’s cloud platform. Citing leaked documents, The Guardian reports that the IDF used Amazon’s Rekognition software to analyze images of Palestinians.

This piece takes no position on whether the IDF’s reported use of Rekognition is right or wrong. The problem is that The Guardian builds its report around a tool it never explains.

The report’s central finding is that the IDF uploaded 3.5 million photographs of Palestinians to Amazon’s cloud infrastructure also known as AWS. It also examines AWS’s broader relationship with the Israeli military, stating that “the ties between AWS and the Israeli military deepened during a period in which Israel has been accused of committing grave crimes in Gaza.” Its own reporting, however, traces the relationship to Project Nimbus, a multibillion-dollar cloud contract Israel signed with AWS and Google in 2021. The Guardian also reports that AWS opened data centers in Israel in early 2023 and that the IDF department known as Rolling Stone moved significant volumes of data into them throughout that year. That migration began months before Hamas and other Palestinian terrorist groups launched the war with the horrific October 7 attacks.

The report surrounds these facts with material that goes beyond the leaked documents. It notes that a UN commission of inquiry, Israeli human rights groups, and genocide experts support the charge that Israel committed genocide in Gaza, a charge Israel denies. It twice mentions NAZA, a Guardian-produced documentary about what the article calls “the remote surveillance and killing of Palestinian civilians,” and it closes by inviting readers to sign up to watch the film. Both bylined reporters worked on NAZA: Harry Davies as a producer and Yuval Abraham as its co-director. The Guardian discloses this only in a note at the bottom of the article.

While the Guardian’s report cites multiple reports and unnamed sources, it leaves significant gaps. The piece does not discuss how much facial matching contributed to identifying a suspect or what additional evidence the IDF required before acting. The Guardian reports that Israel used Rekognition to identify the hostages Hamas and its allies took on October 7, but it does not say how well the system performed. It separately reports that the military used “facial recognition technology” to identify suspected terrorists, without saying whether that technology included Rekognition. Did the IDF treat Rekognition results as investigative leads that required independent verification, or as sufficient grounds for identification and detention? That distinction matters when assessing both the technology’s usefulness and the risks of its deployment.

The distinction matters all the more because the misidentifications the Guardian cites involved a different system. The Guardian points to a New York Times report that facial recognition in Gaza at times wrongly flagged civilians as “militants.” That report covered a program that the Tel Aviv-based company Corsight built and that Israeli military intelligence used alongside Google Photos. It did not involve Amazon’s Rekognition.

Amazon’s own documentation explains how Rekognition works. The software locates a face in an image, maps its features, and compares them with a second face to produce a similarity score. Amazon states that even a 99 percent score does not guarantee a positive match. For public-safety uses, Amazon recommends a threshold of 99 percent or higher and tells users to have human reviewers verify results before they make decisions.

Open-source research shows how the tool works in practice. The author of this article used Rekognition to help identify three October 7 perpetrators. They are linked to the murder of Joshua Mollel at Kibbutz Nahal Oz, after which terrorists took his body to Gaza; the abduction of Raz Ben Ami from Kibbutz Be’eri; and the abduction and murder of German-Israeli national Shani Louk. In each case, reverse-image search tools surfaced a candidate before Rekognition compared faces and returned similarity scores between 99.4 and 99.9 percent. The software corroborated leads that hours of research had already produced. It did not produce those leads. The author published this research at FDD and later in FDD’s Long War Journal. These cases show how Rekognition functions. They say nothing about how the IDF used it.

This is the context the Guardian leaves out. The report never claims that Rekognition delivers a positive match, but it also never explains that the software produces only a similarity score, which Amazon itself says requires human review in public-safety settings. Rekognition is a tool that supports evidence, not a magic bullet. Without that explanation, readers cannot judge what allegedly uploading 3.5 million photos to the platform actually means, or what the IDF could and could not learn from it.


© Divuach Ne’eman R”A, 2025

Joe Truzman

Joe Truzman

Joe Truzman is an open-source intelligence analyst specializing in Palestinian armed groups, Hezbollah, and Iranian-backed terrorist organizations. His work focuses on identifying militant networks, personnel, and emerging security developments across the Middle East.

Discover More

Get clear, reliable Israel news and tools to fight media bias.

Subscribe