Israel put 3.5 million Palestinian photos on Amazon cloud for AI surveillance, leaked files reveal

Leaked documents reveal Israel uploaded 3.5 million Palestinian photographs to Amazon’s cloud for AI facial-recognition analysis, raising questions over military surveillance and civilian privacy.

RNA Media illustration for representation.

New Delhi: The Israeli military transferred approximately 3.5 million photographs of Palestinians to Amazon’s cloud infrastructure for facial-recognition analysis, exposing the growing role of American technology companies in Israel’s surveillance operations in Gaza and the occupied West Bank. The disclosure, based on leaked documents and accounts from military insiders, raises fresh questions about the collection of civilian biometric information, the use of artificial intelligence in military operations and the safeguards governing commercially supplied surveillance technology.

According to an investigation published by the Guardian on Monday, the photographs were uploaded in early 2024 by an Israeli military department responsible for maintaining extensive databases on Palestinians. The department intended to analyse the images using Amazon Rekognition, an AI-powered system capable of detecting faces and comparing photographs to identify possible matches.

The operation took place as Israel intensified its military campaign in Gaza following the Hamas-led raids of October 7, 2023, and expanded surveillance at checkpoints across the territory. The investigation suggests that the Israeli military’s growing dependence on commercial cloud computing extended beyond conventional data storage into sensitive intelligence processing and biometric identification.

The unit behind the transfer is commonly known as Rolling Stone, after a longstanding military information system containing personal records of Palestinians in Gaza and the West Bank. Its databases reportedly include photographs, family relationships, residential and property information, employment permits, checkpoint movements and records relating to suspected criminal activity.

The military department had accumulated millions of photographs through population registration and identity-document systems, including images taken when Palestinians received identification cards at the age of 16. According to the investigation, these records formed part of a broader effort to consolidate information that could be accessed and analysed by Israeli security personnel.

How Amazon’s tech entered surveillance operation

The 3.5 million images were transferred to Amazon Web Services (AWS) infrastructure in Israel, where the military sought to use Rekognition to analyse its photographic databases. Several sources familiar with the operation said the department was experimenting with facial-recognition capabilities as the military expanded its use of automated identification systems.

Facial recognition works by analysing distinguishing features of a face and comparing them with photographs held in an existing database. According to Amazon’s technical documentation, Rekognition can identify faces in images, compare facial characteristics and generate similarity scores indicating whether two photographs may show the same individual.

The distinction is important because a similarity score does not establish a person’s identity with certainty, particularly when images are blurred, poorly illuminated or captured from unusual angles. Amazon’s own developer guidance acknowledges the possibility of incorrect matches and recommends human verification before facial comparisons influence decisions affecting individual rights or access to services.

Israeli forces reportedly collected photographs of Palestinians passing through temporary military checkpoints established during the Gaza offensive, including people displaced by fighting and evacuation orders. These images were compared against existing records, although investigators could not establish whether photographs collected at those checkpoints were included in the 3.5 million files transferred for Rekognition analysis.

Military sources nevertheless indicated that Amazon’s cloud infrastructure supported some checkpoint-related facial-recognition activities because of the enormous volume of images being processed. The investigation also found that a small group of Israeli soldiers worked from AWS offices in Tel Aviv in early 2024 to explore improvements to the military’s surveillance capabilities.

AWS maintained that visits by customers to its offices were routine business engagements and rejected suggestions that such meetings demonstrated unusual involvement in military operations. The company also said it lacked visibility into customers’ cloud workloads and would not discuss individual commercial relationships without their consent.

Project Nimbus and military’s expanding cloud infrastructure

The revelations draw attention to Project Nimbus, the multibillion-dollar agreement awarded in 2021 under which Amazon and Google provide cloud computing services to Israeli government institutions, including the defence establishment. The programme has enabled Israeli agencies to shift large quantities of information from internally operated systems to commercial data centres offering advanced processing capacity and artificial intelligence services.

Documents examined by the Guardian indicate that AWS provided the Israeli military with a segregated and specially secured cloud environment described as a “walled garden”. The facility was designed to accommodate sensitive military information, including intelligence records, photographs and videos, with substantial data transfers already underway during 2023.

The Guardian previously reported on contractual provisions that allegedly restrict Amazon and Google’s ability to suspend Israeli government and military services even when their terms of use are breached. Such arrangements could considerably limit the technology providers’ discretion to withdraw infrastructure subsequently linked to controversial surveillance operations.

The issue acquired greater significance after Microsoft restricted an Israeli military intelligence unit’s access to cloud and AI services in September 2025. The decision followed an investigation into the use of Microsoft’s Azure infrastructure to store intercepted Palestinian telephone conversations on a massive scale, demonstrating that commercial technology providers can face direct questions about their responsibilities for military surveillance.

Earlier surveillance systems and wrongful identification

The latest findings reinforce concerns raised by Amnesty International in 2023, when it documented the deployment of Israeli facial-recognition systems known as Red Wolf, Blue Wolf and Wolf Pack in the occupied West Bank. Amnesty alleged that these technologies enabled extensive monitoring of Palestinians and contributed to restrictions on movement, particularly at military checkpoints in Hebron.

The use of AI-assisted identification has also extended to operations against suspected participants in the October 2023 Hamas attacks, as RNA Media reported in August. Separately, an RNA Media report on the documentary “Naza” examined Israeli military testimony about AI-assisted targeting, civilian casualties and human oversight during the Gaza conflict.

The central concern is whether automated identification systems can distinguish reliably between civilians and security suspects when their findings influence detention or military decisions. Previous reporting has documented cases in which Israeli facial-recognition technology incorrectly identified Palestinians as suspected militants, highlighting the potentially serious consequences of errors in conflict environments.

Israel maintains that its intelligence and security operations are intended to counter terrorism and protect its population, but its military declined to discuss the operational arrangements described in the latest investigation. The Israel Defence Forces said it would not comment on services connected with operational requirements, leaving unanswered questions about how extensively Amazon’s technology was used and what safeguards governed its deployment.

The leaked documents establish that millions of Palestinian photographs were uploaded for facial-recognition analysis, but they do not independently establish how many operational identifications resulted from that particular transfer. The distinction matters because the broader issue now extends beyond Israel’s surveillance practices to the responsibility of technology companies whose infrastructure makes large-scale military data collection and automated identification possible.

Exit mobile version