TL;DR: Leaked documents obtained by The Guardian, +972 Magazine, and Local Call reveal that ICE more than tripled its Microsoft Azure data storage (from 400TB to nearly 1,400TB) between July 2025 and January 2026. ICE spent $38 million on Microsoft software during this period, with access to Azure AI Video Indexer and Azure Vision tools that can analyze images, detect faces, and identify emotions. Microsoft's official response: they "do not believe" ICE is using the tech for mass surveillance. They also admit they have no visibility into what ICE actually stores on their servers.
The Data Explosion
In July 2025, ICE stored 400 terabytes of data on Microsoft Azure servers. By January 2026, that number hit 1,400 terabytes: a 250% increase in six months.
To put that in perspective: 1,400 terabytes could hold roughly 700 million high-resolution photos, or 350,000 hours of surveillance video, or the biometric data of hundreds of millions of people. We don't know which it is. Neither does Microsoft.
The leaked documents, first reported by The Guardian on February 17, 2026, show ICE's spending on Microsoft technology ramped up dramatically between March and November 2025. The agency paid $38 million for Microsoft software through Dell Federal Systems, a reselling partner. That's on top of $25 million for Amazon Web Services cloud systems and $530,000 for Google Cloud services.
What AI Tools ICE Has Access To
The leaked documents show ICE is using Microsoft's AI video analysis suite. These tools include:
- Azure AI Video Indexer: Analyzes video files to extract faces, emotions, objects, spoken words, and text visible in footage
- Azure Vision: Image analysis including face detection, emotion recognition, and object identification
- Azure cognitive services: Broader AI toolkit for processing and analyzing media
These capabilities align with ICE's known surveillance methods. The agency already uses Palantir's ELITE system to generate "confidence scores" for targeting individuals, Zignal Labs to monitor 8 billion social media posts, and a mobile app called Fortify that scans faces against a 1.2 billion image database in seconds.
Azure's AI tools would complement this stack. Run surveillance video through Video Indexer, extract faces, cross-reference against biometric databases, flag individuals for enforcement. The infrastructure supports it.
Microsoft's Non-Denial Denial
When asked about the leaked documents, Microsoft issued a carefully worded statement: "Microsoft provides cloud-based productivity and collaboration tools to DHS and ICE, delivered through our key partners. Microsoft policies and terms of service do not allow our technology to be used for the mass surveillance of civilians, and we do not believe ICE is engaged in such activity."
Parse that carefully. Microsoft "does not believe" ICE is doing mass surveillance. Not "ICE isn't doing it." Not "we've verified they're not." Just... they don't think so.
The company also told The Guardian it has "no visibility" into what kind of data ICE actually stores on Azure. So Microsoft's belief that ICE isn't doing mass surveillance is based on... nothing. They don't know what's in those 1,400 terabytes. They can't know.
Their policies prohibit mass civilian surveillance. Their contracts prohibit it. But they can't see what's being stored, they can't audit how the AI tools are being used, and their assurance amounts to "we trust them."
Microsoft Employees Aren't Buying It
The Guardian's reporting includes a detail Microsoft would rather ignore: some employees have filed internal ethics reports expressing concern that the company's technology might be enabling operations involving excessive force or unlawful activity.
This isn't new. Microsoft employees have pushed back on ICE contracts since 2018, when the company's Azure Government relationship with the agency first came under scrutiny. CEO Satya Nadella defended the contracts then, saying they were for "legacy mail, calendar, messaging." Six years later, ICE is storing 1,400TB on Azure and has access to AI video analysis tools.
The internal ethics complaints haven't stopped the contracts. They haven't triggered audits. They've just been filed and, apparently, ignored.
The Bigger Picture
ICE's cloud expansion coincides with a massive ramp-up in enforcement activity. The agency's budget has swelled, its workforce expanded rapidly, and arrest numbers are up 40% compared to January 2025.
The human cost is visible in the numbers. Over 4,400 court rulings since October 2025 have found unlawful immigrant detention. U.S. citizens Renee Good and Alex Pretti died in ICE-related operations. The agency is operating at a pace that courts are repeatedly finding illegal.
All that activity generates data. Surveillance footage from raids. Biometric scans from detention facilities. Social media monitoring logs. Location data purchases. It has to go somewhere. Now we know: 1,400 terabytes of it sits on Microsoft servers.
What We Don't Know
The leaked documents don't specify what ICE actually stores in those 1,400 terabytes. The agency could be using Azure for administrative functions, running detention center logistics, coordinating deportation flights, managing payroll.
But ICE's documented surveillance infrastructure (the Palantir ELITE system, the Zignal social media monitoring, the facial recognition apps, the location data purchases) all generate massive amounts of data. That data has to live somewhere. A platform with AI video analysis tools that can process faces and emotions would be convenient.
We can't prove ICE stores surveillance data on Azure. We can prove they tripled their storage, spent $38 million on Microsoft tools, have access to AI analysis capabilities, and Microsoft has no idea what they're actually doing with it.
What You Can Do
For Affected Communities
- Assume ICE has access to sophisticated AI analysis tools
- Review our ICE surveillance protection guide
- Minimize social media presence and location data exposure
- Connect with immigrant rights organizations in your area
For Microsoft Employees
- Continue filing internal ethics reports. They create a paper trail
- Connect with Microsoft employee advocacy groups
- Document concerns about specific projects
- Consider speaking with journalists (with appropriate protections)
For Microsoft Customers
- Ask Microsoft directly about ICE contract auditing
- Support shareholder resolutions requiring human rights reviews
- Consider whether your organization's cloud choices align with your values
The Bottom Line
Microsoft's position is untenable. They sell ICE sophisticated AI surveillance tools. They store 1,400TB of ICE data. They admit they have no visibility into what that data contains. And they assure us ICE isn't doing mass surveillance because... their contract says so.
Contracts don't enforce themselves. Policies don't audit compliance. And "we don't believe they're doing it" means nothing when you've deliberately structured the relationship to avoid knowing what your customer actually does.
Microsoft isn't just providing cloud services to ICE. They're providing plausible deniability.
References
- +972 Magazine - ICE tripled its reliance on Microsoft in last six months, files reveal (February 2026)
- American Bazaar - Microsoft says it 'does not think' ICE uses its tech for mass surveillance (February 2026)
- Data Center Dynamics - US' ICE triples use of Microsoft cloud in six months (February 2026)
- WinBuzzer - ICE Triples Azure Usage to 1,400TB Using Microsoft's AI Surveillance Tools (February 2026)
- TechCrunch - Here's the tech powering ICE's deportation crackdown (January 2026)
- WBUR - How ICE is using surveillance technology in immigration crackdowns (February 2026)