How Israel used Azure to monitor Palestinians | Explained

The story so far: In August, a joint investigation by The Guardian, +972 Magazine and Local Call revealed that Israel’s military intelligence unit had built a cloud-based surveillance system using Microsoft’s Azure to store audio recordings of Palestinians’ phone calls. Unit 8200, which is considered Israel’s equivalent to the U.S.’s National Security Agency, was reportedly uploading “audio files of millions of calls by Palestinians in the occupied territories” into a dedicated Azure environment.

What was Microsoft’s initial response?

Microsoft’s initial response downplayed any wrongdoing. The company said it had “found no evidence” that its tools were used to harm civilians and claimed to be unaware of how the cloud project was being used. But following the expose, the software giant launched an internal review. And on September 25, the company announced that its review had “found evidence that supports elements” of the reporting and that it had “ceased and disabled a set of services” to the Israeli defence unit involved.

How did this partnership begin?

The project was hatched in late 2021 when Unit 8200’s commander, Yossi Sariel, met with Microsoft CEO Satya Nadella in Seattle. Mr. Sariel pitched to move a huge portion of Israel’s collected intelligence data onto Azure. Mr. Nadella reportedly agreed to a proof-of-concept plan and assigned a dedicated team of Microsoft engineers to work closely with Unit 8200 to build a secure, custom Azure deployment accessible from military bases. In Microsoft’s internal view, this partnership was treated as “critical” to the company and even a “powerful brand moment” for Azure.

The military’s reason was simple: the existing Israeli data centres could not hold vast volumes of intercepted phone traffic. Under Mr. Sariel’s tenure, Unit 8200 had dramatically expanded its surveillance and integrated multiple databases of Palestinians. Previously, the unit could keep the calls of only some tens of thousands of individuals on its own servers. By moving to Azure, it suddenly had “infinite” storage. In effect, once the cloud project was approved, Unit 8200 could funnel nearly all phone intercepts into Azure and scale its watchlist from thousands of calls per day to millions.

How did the surveillance work?

The mechanics of the system hinged on cloud computing. Essentially, Unit 8200’s field listening posts and telecom intercepts were piped into a dedicated Azure “instance”. Microsoft and IDF engineers built a segregated Azure environment so that data from Gaza and the West Bank could be uploaded continuously and stored long-term. Because Azure offers elastic capacity and advanced AI tools, the unit could automate analysis on this content.

According to published reports, all intercepted voice calls and texts were transcribed and translated in the cloud. Advanced search indexes were then used to analyse the data to quickly decipher patterns. In short, Azure acted as both an enormous archive and an AI-driven search engine. The Guardian’s investigation emphasised that the IDF’s calls database was kept in “a customised and segregated area within the Azure platform,” allowing it to be “analysed using AI-driven techniques”. This kind of setup is enabled by modern cloud features that are powered by multi-modal AI that can turn any data into searchable text. Once a phone call is converted by speech-to-text, it is treated like any document. In practice, such tools would allow an intelligence officer to input a query and quickly retrieve relevant conversation snippets or even link them to satellite imagery.

Where did phone usage data come from?

Crucially, none of this would have worked if Palestinian networks were independent. While, under the 1995 Oslo accords, Israel recognised Palestinians’ right to build and operate their own communications systems, including mobile networks, in practice, Israel retained control over almost all telecommunications infrastructure in Gaza and the West Bank. That means virtually every Palestinian cellphone call and data packet passes through Israeli-controlled switches.

As The Guardian’s report notes, “thanks to the control [Israel] exerts over Palestinian telecommunications infrastructure, Israel has long intercepted phone calls in the occupied territories”. Without an independent mobile backbone, Palestinians must use cellular services that are extensions of Israeli operators, so mass monitoring of local networks has long been routine. Multiple analyses stress that Israel’s continued monopoly or de facto control of Palestinian internet and phone lines violates the spirit of Oslo’s “separate infrastructure” provisions. Israeli bureaucrats often justify such surveillance on security grounds, but rights advocates warn that it amounts to mass spying.

Will the surveillance stop?

Microsoft’s Azure was not the only cloud in use. Earlier investigations revealed that the Israeli military also tapped Amazon Web Services (AWS) for some of its data storage. Moreover, the IDF has a broader cloud strategy: it signed a $1.2 billion deal, nicknamed “Project Nimbus”, under which both Google Cloud and Amazon provide computing and AI services to the military. Even IBM subsidiary Red Hat has supplied cloud infrastructure, and Palantir, a U.S. defence tech firm, offers AI targeting tools to Israel. In short, besides Microsoft, almost all major U.S. tech firms have become entwined with Israel’s digital war effort. What this means is that cutting off Azure does not leave Unit 8200 defenceless; it will simply shift workloads to AWS, Google or a combination of on-premises servers.