Google employees, kept in the dark about firm's sale of advanced AI tools to Israel, say it could bolster Israel's ongoing military occupation of Palestine
Sam Biddle / The Intercept : Tweets: @timnitgebru , @amnestytech , @jathansadowski , and @samfbiddle Tweets: Timnit Gebru / @timnitgebru : “documents indicate that the new cloud 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.” https://theintercept.com/... @amnestytech : “Google's machine learning capabilities along with the Israeli state's surveillance infrastructure poses a real threat to the human rights of Palestinians,” says @amnesty's @Damini_Satija https://twitter.com/... Jathan Sadowski / @jathansadowski : “Sadowski says that some technologies amplify the exercise of power to such an extent that even their use by a country with a spotless human rights record would provide little reassurance.” I spoke with @samfbiddle for his excellent, in-depth piece: https://theintercept.com/... Sam Biddle / @samfbiddle : NEW: Google training documents show the company's “Project Nimbus” contract with the Israeli government and military includes a variety of AI and ML tools, technology critics worry will worsen the surveillance of Palestinians and bolster the occupation https://theintercept.com/...
Context & Ripple Effects
This 2022 report is the opening document in what became a years-long arc around Project Nimbus, Google's $1.2B cloud-and-AI contract with Israel. Employees say the sale of facial detection, object tracking, and sentiment-analysis capabilities was made without internal disclosure — a sharp contrast with 2020, when employee outcry ended Google's Maven work, even as the company was aggressively pursuing a major Pentagon cloud contract and Google Cloud AI had already surfaced in CBP's border-surveillance stack with Anduril.
Amnesty's tech team frames the combination of Google's ML and Israel's surveillance infrastructure as a direct human-rights threat to Palestinians. The later record — Google's own lawyers worrying in early 2021 that Nimbus could enable rights violations, a 2023 staff open letter over pro-Palestinian speech limits, and documents showing Israel got access to Google's latest AI tools in the war's early weeks — makes this the origin point of the controversy.
First-order effects
- Google's own workforce learns about a strategic government AI sale from leaked documents rather than internal disclosure, deepening the trust gap that already drove the Maven walkout and later the staff open letter over speech limits.
- Amnesty's tech team and allied researchers gain documentary evidence — facial detection, automated image categorization, object tracking, sentiment analysis — to argue the contract directly expands surveillance capabilities over Palestinians.
Second-order effects
- The disclosure forces Google to manage Nimbus as a reputational and governance liability: the company later drafts an adjusted terms-of-service with Israel rather than applying its general ToS, and internal legal and policy teams document their own human-rights concerns.
- Rival cloud vendors competing for government AI work face the same dual-use exposure, while activist and employee scrutiny becomes a standing cost factor in any state surveillance contract Google or its peers sign.
Third-order effects
- If the Nimbus pattern holds — commercial AI labs selling advanced capabilities to governments under bespoke terms, with employees excluded — the industry structurally reverts to the defense-contractor model the Maven protest was meant to prevent, with dual-use governance decided by contract language rather than internal review.
The trend: Frontier AI labs are becoming state surveillance suppliers under bespoke government contracts, with employee oversight and human-rights review steadily losing ground to deal-making.