JupiteX Get the app
Defence & Security16 Nov 2026 · about 5 min

IDF uploaded 3.5 million Palestinian photos to Amazon servers for facial recognition, report claims

The brief

The investigation alleges that an IDF department uploaded millions of photographs of Palestinians to Amazon’s cloud servers. The images were intended for a facial recognition project using Rekognition, Amazon’s AI-powered image analysis platform. The report matters because the photographs included sensitive information about Palestinians, and facial recognition can be used to identify people from images.

01

What does the investigation allege the IDF uploaded to Amazon’s servers, and what was it intended to do with the images?

The investigation alleges that an IDF department uploaded millions of photographs of Palestinians to Amazon’s cloud servers. The images were intended for a facial recognition project using Rekognition, Amazon’s AI-powered image analysis platform. The report matters because the photographs included sensitive information about Palestinians, and facial recognition can be used to identify people from images.

02

How many photographs were reportedly uploaded, and when and where were they collected or stored?

The reported scale was 3.5 million photographs of Palestinians, uploaded by Israel’s military to Amazon cloud servers. The number matters because the images were part of a facial-recognition effort, not simply an ordinary photo archive. The Guardian’s investigation based its account on leaked documents and interviews with military sources.

According to the report, the IDF collected images at checkpoints after October 7 attacks, including people displaced by airstrikes. The upload occurred during the war’s early months; Amazon was notified in early 2024 about planned Rekognition use. The data reportedly went to an Amazon data center near Tel Aviv, as part of wider transfers to AWS.

The report says Rekognition helped identify hostages and suspected militants; earlier tests reportedly misidentified civilians. The IDF declined to comment on operational contracts or services. Amazon said it lacked visibility into customer workloads and would not discuss customer relationships without consent.

03

What is facial recognition, and how is Amazon Rekognition used to identify people in images?

Facial recognition software analyzes a face in a photo or video to help identify or compare someone. It can match facial patterns against images already stored in a database. This matters because a match can influence whom authorities investigate or seek.

In early 2024, the IDF reportedly sent 3.5 million Palestinian photographs to Amazon servers near Tel Aviv. The investigation said the military planned to use Rekognition, Amazon's AI image-analysis service, for facial recognition. During the war's early stages, the IDF reportedly used Rekognition to identify hostages and people suspected of terrorism.

The report did not state how often the system was accurate. The New York Times had reported that similar commercial tools sometimes misidentified civilians as terrorists, highlighting the importance of accuracy and human review. Amazon said it could not see customers' workloads, while the IDF declined to comment on operational services.

04

What can happen when facial recognition misidentifies someone, and what errors were reported in Gaza?

Facial recognition matches a person's face against stored images; an incorrect match can wrongly mark an innocent person as a suspect. That matters because officials may act on the label, potentially bringing scrutiny or danger to someone misidentified. The article reports false matches, but does not describe specific consequences for those civilians.

During the early Gaza war, the IDF reportedly used Amazon Rekognition to identify hostages and people suspected of terrorism. The New York Times reported Gaza experiments identifying people in crowds and drone footage; civilians were misidentified as terrorists. A mistaken match could turn an ordinary civilian into a false lead; the reporting gives no case-level outcomes.

Guardian investigators say the IDF uploaded 3.5 million Palestinian photos to an Amazon center near Tel Aviv. The article also says photos were collected at checkpoints, including from people displaced by airstrikes. It gives no accuracy rate or proof every image was used, leaving error risks and impacts unclear.

05

What does it mean to store sensitive information in a commercial cloud service such as Amazon Web Services?

A commercial cloud stores information on a provider's computers, instead of only on the customer's own systems. Customers can use online storage and tools, while setting access permissions and security protections. This matters because sensitive data may be handled outside the organization's direct control, so access rules and safeguards are important.

The Guardian reported that the IDF sent Amazon 3.5 million Palestinian photographs during the war's early months. It said the IDF notified Amazon in early 2024 that it wanted to use Rekognition for facial recognition. The report described cloud storage as part of a project to analyze images, not simply keep a backup.

AWS said it had no “visibility” into customers’ cloud workloads and declined to discuss its relationship without consent. The IDF said it does not comment on contracts or services tied to operational needs. For sensitive information, the practical questions are who can reach it, what protections apply, and who can explain its use.

06

How do facial recognition systems compare faces, and why can the results be inaccurate?

Facial recognition systems turn visible facial patterns into a digital template, then compare it with stored templates. They estimate similarity, so a match is a lead, not certain proof of identity. That distinction matters when authorities use results to make decisions about people.

The Guardian reported that Israel’s military uploaded 3.5 million Palestinian photographs to Amazon servers. It reportedly notified Amazon in early 2024 that it wanted Rekognition for facial recognition. The New York Times had reported that tests in crowds and drone footage sometimes misidentified civilians as terrorists. Poor images, unusual angles, and similar-looking faces can also cause errors; those causes were not identified in the report.

The investigation says soldiers collected checkpoint photos, including images of people displaced by airstrikes. Amazon said it could not see what customers ran in its cloud. Because a false match can implicate someone, recognition results need human checks and limits, especially in conflict zones.

This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.

Read more in the JupiteX app

Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.

Or read more news on the web