Revealed: Israel uploaded millions of photos to Amazon’s facial recognition platform to track Palestinians
The Israeli military uploaded 3.5m photos of Palestinians to Amazon’s cloud platform, according to documents obtained by the Guardian. The upload was intended to support a facial-recognition project using Amazon Rekognition, an AI-powered image-analysis service. The scale matters because the collection covers Palestinians in Gaza and the West Bank, and the article says such personal data was often gathered without consent. The files show that a military department known as Rolling Stone moved the images into an Amazon datacentre near Tel Aviv in early 2024. The department holds a broad catalogue of images and was exploring facial-recognition tools to analyze it. A military source said Amazon’s cloud infrastructure was used for the very large image volumes collected at checkpoints, but the article does not establish that these were the 3.5m uploaded photos. The military’s precise purpose for each image, and whether the project became operational, remain unclear. An AWS spokesperson said the company had no visibility into customers’ workloads and would not discuss its relationship without consent. The documents nevertheless show the military’s use of Amazon infrastructure for image analysis and storage.
How many photos of Palestinians did the Israeli military upload to Amazon’s platform, and what was it trying to do with them?
The Israeli military uploaded 3.5m photos of Palestinians to Amazon’s cloud platform, according to documents obtained by the Guardian. The upload was intended to support a facial-recognition project using Amazon Rekognition, an AI-powered image-analysis service. The scale matters because the collection covers Palestinians in Gaza and the West Bank, and the article says such personal data was often gathered without consent.
The files show that a military department known as Rolling Stone moved the images into an Amazon datacentre near Tel Aviv in early 2024. The department holds a broad catalogue of images and was exploring facial-recognition tools to analyze it. A military source said Amazon’s cloud infrastructure was used for the very large image volumes collected at checkpoints, but the article does not establish that these were the 3.5m uploaded photos.
The military’s precise purpose for each image, and whether the project became operational, remain unclear. An AWS spokesperson said the company had no visibility into customers’ workloads and would not discuss its relationship without consent. The documents nevertheless show the military’s use of Amazon infrastructure for image analysis and storage.
What is Amazon Rekognition, and how does facial recognition use a photo to try to identify someone?
Amazon Rekognition is an AI-powered image-analysis platform. The article says the Israeli military wanted to use it for a facial-recognition project involving a large collection of Palestinian images. It does not explain Rekognition’s technical design or say how the project made its matches.
In general, facial-recognition systems locate a face in an image and measure visual patterns, such as the spacing and shape of facial features. They turn those measurements into a digital representation, then compare it with representations from other images. The system may return likely matches or a similarity score, rather than independently proving who someone is. These are general details about the technology, not specifics reported about the military’s project.
Results can depend on image quality, lighting, camera angle, and how well the system performs for different people. A match can be wrong, so it should not be treated as certain identity without further checks. The article notes previous reporting that commercially available facial-recognition technology used in Gaza wrongly flagged some civilians as militants, underscoring the stakes of relying on automated matches.
Where and how has the Israeli military collected images of Palestinians?
The article describes image collection in both the West Bank and Gaza. Israel has long used facial recognition in the occupied West Bank, where face scanning is routine at checkpoints. Israeli soldiers also use custom-built smartphone apps to collect images of people living in Palestinian territory.
After the ground invasion of Gaza began in October 2023, the military routinely collected images of Palestinians as they passed through checkpoints. Many had been forcibly displaced by the bombing campaign. The checkpoints could act as temporary surveillance posts. A military source familiar with the effort said Amazon’s cloud infrastructure was used because of the very large volumes of images collected at these sites.
However, the article says it is not clear whether the 3.5m photos uploaded to Rekognition came from those checkpoints. The images were held by a military department with a vast catalogue covering Palestinians in Gaza and the West Bank. The source text does not give a complete account of when or where every image in that catalogue was captured.
What can happen when facial recognition wrongly identifies a civilian as a militant?
Facial recognition can make mistakes, and the consequences may be serious when a wrong match influences military decisions. The article says technology used in Gaza sometimes wrongly flagged civilians as militants. It does not report what happened to those individuals or establish that a particular person was harmed because of a false match.
In general, if officials rely on an incorrect identification, a civilian could face questioning, detention, or other scrutiny. In a conflict setting, a mistaken label could also contribute to dangerous decisions about how that person is treated. These are potential consequences of a false identification, not outcomes documented for the civilians mentioned in the article.
The article describes experiments using commercially available technology to detect faces in crowds and drone footage. It also says the military used facial recognition to identify people it suspected were militants. When automated results feed into those kinds of decisions, human review and checks matter; a computer-generated match should not be treated as proof. The source does not describe safeguards used in the military’s projects.
What is Project Nimbus, and how does it connect Amazon and Google’s cloud services to Israeli government and military work?
Project Nimbus is a cloud-computing procurement project involving Amazon Web Services and Google. As outside context, it is widely described as a contract to provide cloud services to the Israeli government and its defense establishment. The article excerpt itself does not name Project Nimbus, so it does not directly connect the 3.5m-photo upload to that project.
The article does describe a broader relationship: the Israeli military moved large amounts of data into Amazon datacentres and used AWS infrastructure for image analysis. It reports that a small group of Israeli soldiers worked from AWS’s Tel Aviv offices in early 2024 to explore ways to support a facial-recognition programme in Gaza. AWS described customer visits as routine engagement.
The distinction matters: Nimbus is the wider government cloud context, while the documents discussed in the article concern military use of Amazon’s services. The excerpt does not explain the terms of Nimbus, identify Google’s role in this specific military work, or establish that the uploaded photos were stored under the Nimbus contract. Those links should not be assumed.
Why can storing sensitive personal data in a commercial cloud raise privacy concerns, especially when people did not consent to its collection?
Sensitive personal data can reveal where people live, how they move, or who they may be connected to. Storing it in a commercial cloud raises questions about who can access it, how it is protected, how long it is retained, and what it may be used for. These are general privacy concerns; the article does not report that Amazon misused or exposed the data.
The documents describe the Israeli military moving large volumes of intelligence and personal information about Palestinians into Amazon datacentres. The article says this information was typically gathered indiscriminately, through intrusive means and without people’s consent. Combining a large photo collection with facial recognition can make it easier to search and analyze images at scale, increasing the importance of clear limits and safeguards.
AWS said it had no visibility into the workloads customers hold in its cloud and would not discuss its relationship without customer consent. That response does not resolve questions about the military’s collection or use of the data. The article highlights how cloud services support that work, while leaving unclear the full contents of the uploaded collection and whether the facial-recognition project became operational.
What is cloud computing, and how does it let organizations store and analyze very large collections of data?
Cloud computing means using computing services delivered over a network, usually the internet, rather than relying only on computers and servers an organization owns and operates itself. Those services can include data storage, computing power, and software tools. This definition is general background; the article focuses on Amazon’s cloud infrastructure and Rekognition.
Organizations can send large collections of files to datacentres, where they can be stored and processed using connected computing resources. They can then use analysis tools to search or examine the data without keeping every file on a local device. The article says the Israeli military moved 3.5m photos into an Amazon datacentre near Tel Aviv and used AWS infrastructure because of the very large image volumes involved.
Cloud capacity can help organizations handle data at a scale that would be difficult with a small local system. But moving data to a cloud does not answer who may access it or how it will be used. In this case, AWS said it had no visibility into customer workloads, while the documents raised questions about sensitive personal data stored and analyzed through its services.
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