What are the Project Nimbus protests really about?
Earlier this month, employees of Google and Amazon took to the streets to protest their employers’ work with the Israeli government through Project Nimbus, a cloud computing arrangement that was announced in early 2021. At one rally, a Google software engineer described Project Nimbus as an attempt on Google’s part to “try and become a military contractor,” while another organizer urged the crowd “to fight back, to make sure the technology we build is for good.”
Facial recognition technology is, of course, part of the issue. This summer, The Intercept reported that secret documents revealed that Google is providing the Israeli government with a cloud system that “would give Israel capabilities for facial detection, automated image categorization, object tracking, and even sentiment analysis that claims to assess the emotional content of pictures, speech, and writing.”
The report prompted action from at least one Google employee. Ariel Koren, a marketing manager for Google’s educational products, spent over a year trying to organize fellow employees against Project Nimbus, and ultimately resigned in protest. In her public resignation letter, she cited the above text from The Intercept to argue that it “is clear that the tools provided through Nimbus have the potential to expand Israel’s pattern of surveillance, racial profiling, and other forms of tech-assisted human rights violations.”
Koren succeeded in attracting media attention to the issue, with The New York Times reporting about her efforts to protest a “$1.2 billion agreement for Google and Amazon to supply Israel and its military with artificial intelligence tools and other computing services.”
But what exactly is Project Nimbus? Various media reports have described the program as being shrouded in mystery, while Google itself has framed it in banal, enterprise IT terms.
The Tell-Tale Slide
When it publicly announced the project last year, Israel’s Finance Ministry said that Amazon and Google had been “selected as the providers awarded the government’s cloud services,” adding, “The project is intended to provide the government, the defense establishment and others with an all encompassing cloud solution.”
Responding to Gizmodo’s inquiries a year and a half later, a Google spokesperson left out the “defense establishment” bit. “As we have stated many times, the contract is for workloads running on our commercial platform by Israeli government ministries such as finance, healthcare, transportation, and education,” the spokesperson said, adding that the employees protesting the project were “misrepresenting the contract—our work is not directed at highly sensitive or classified military workloads relevant to weapons or intelligence services.”
That response would seem to rule out the provision of facial recognition for surveillance purposes; but that would seem to contradict The Intercept’s reporting.
To help clarify the matter, it’s worth taking a closer look at the source materials cited by The Intercept, whose report helpfully links to at least some of the training documents it obtained with respect to Nimbus. So far as facial recognition technology is concerned, the reporting appears to be based on one page from a 179-page slide deck titled “Welcome to Cloud OnBoard”, and detailing the capabilities of Google Cloud. Page 100 lists a number of the capabilities of the platform’s “Cloud Vision API”, including things like computer vision-powered face detection, object labelling, and Optical Character Recognition.
In other words, Google Cloud supports these capabilities as part of its overall platform; they haven’t been tailored for the Israeli government’s surveillance needs. As The Intercept itself noted in its report, for the most part the documents aren’t unique to Nimbus; rather, they “appear to be standard educational materials distributed to Google Cloud customers and presented in prior training contexts elsewhere.”
Does that definitely rule out the Israeli military’s use of Google’s facial recognition technology? No. But why would the Israeli military – one of the most technologically sophisticated in the world – need Google’s “facial detection” technology when it already has Blue Wolf?
A Pack of Wolves
As detailed in a Washington Post feature last November, Blue Wolf is a biometric surveillance system that the Israeli military had been rolling out since 2019, and is designed to let soldiers use a smartphone camera to match Palestinian faces against a comprehensive database. One former soldier called it the army’s “Facebook for Palestinians.” A former soldier also told the Post that the system is a pared-down version of a larger, restricted-access system called Wolf Pack, which “contains profiles of virtually every Palestinian in the West Bank,” according to the report.
This is the kind of thing that makes privacy and human rights advocates very concerned – a profoundly invasive surveillance system that would not be tolerated in Israel itself, not to mention many of the US jurisdictions that have sought to curtail the use of biometric surveillance by police in recent years. But there is no suggestion that Google had played a secret role in its construction before winning the Nimbus contract, and there’s no reason to think that the Israeli military would see Google’s computer vision technology as an important addendum to its capabilities now. In fact, when the Israeli government has needed facial recognition technology in the past, it has turned to biometrics specialists, rather than to a computer vision generalist like Google.
Indeed, for many protestors, it may be that Google’s actual involvement (or lack thereof) in such surveillance programs is almost beside the point – that the real issue is Google’s having any business with a government that would engage in such practices. As one Amazon worker involved in the recent protests explained to Gizmodo, “There is no way for Amazon and Google to justify a contract with a government that has violated numerous human rights and continues to oppress Palestinian lives.” Meanwhile, the aforementioned Google engineer who had told protestors that Google was trying to become a military contractor added, “Please help us in keeping Google from becoming complicit from apartheid.”
Such arguments can be – and have been – made without the need for spurious claims about Google or Amazon providing AI technology to the Israeli military. But the invocation of facial recognition in the protest coverage helps to illustrate another way this technology can be weaponized – as a rhetorical tool. Facial recognition has, in recent years, become intertwined with controversies around policing and intrusive surveillance all of the world. For many, tying it to an issue of concern offers a way of immediately making it seem more sinister.
That having been said, important questions about Project Nimbus remain. It still isn’t clear why Israel’s Ministry of Finance made special mention of the project’s support for “the defense establishment”, separate from the government, in its initial announcement. And, curiously, little attention has been paid to Google’s partner in the project, Amazon, which really has developed its own bona fide facial recognition system, and has sold it to police agencies in the US in the past. There is no evidence – not even a slide deck – hinting that Amazon is selling its Rekognition system to Israel via the Nimbus contract, but that doesn’t rule out the possibility.
It may turn out that Nimbus’s cloud technology will merely support the Israeli defense establishment’s payroll and accounting operations, just as it potentially will across the rest of the Israeli government’s arms. Or it may transpire that Nimbus does end up delivering biometric support services to the military. But the latter probably won’t come from Google, and government officials won’t be learning about it through Slide 100 of the Google Cloud training deck.