When government surveillance goes digital, it doesn’t just look through your window – it scrolls your feed. 

Immigration and Customs Enforcement (ICE) has quietly signed a $5.7M deal for AI-driven social media surveillance software. 

The technology, developed by a Silicon Valley firm called Zignal Labs and distributed by Carahsoft Technology, promises to monitor over 8B posts a day. 

This isn’t a one-off experiment. It’s a five-year contract, giving ICE’s intelligence unit, Homeland Security Investigations, real-time access to a platform originally built for PR firms and political campaigns. 

The same software that once helped brands track hashtags is now being used by law enforcement to find ‘threats.’

What exactly qualifies as a ‘threat,’ of course, is where things get interesting.

The Deal: Zignal Labs Joins the ICE Toolkit

The September procurement notice is short on details, but the paper trail is clear. 

Zignal Labs, a data analytics company founded in 2011, has quietly shifted from monitoring brand sentiment to supplying tactical intelligence to the Pentagon, the Israeli military, and now ICE.

The pitch is simple: Zignal’s AI scans social platforms, aggregates billions of data points, and delivers ‘curated detection feeds’ so investigators can ‘respond to threats with greater clarity and speed.’ 

The government calls that situational awareness. Privacy advocates call it mass surveillance.

The Department of Homeland Security has used Zignal before – the Secret Service was the first to get licenses back in 2019. 

But this is the first known deal that places the software directly in ICE’s hands. 

It adds another layer to an already complex surveillance network, which includes ShadowDragon (that maps online activity) and Babel X (that links social media profiles to real-world identifiers, such as Social Security numbers).

Together, these tools give ICE a nearly panoramic view of digital life – one that can easily extend beyond immigration enforcement into political monitoring.

Building the AI Surveillance Infrastructure

The ICE-Zignal deal isn’t happening in isolation. It’s part of a broader, well-funded trend: government agencies adopting AI tools from private defense tech firms.

In 2021, Zignal announced its new ‘public sector advisory board’ and a pivot toward military and intelligence clients. 

In one brochure, the company boasted of giving ‘tactical intelligence’ to ‘operators on the ground’ in Gaza – the same tech now wired into U.S. domestic policing.

In July, Zignal partnered with Carahsoft Technology, a federal IT contractor that distributes a range of solutions, including Splunk dashboards and Palantir-adjacent analytics. 

The new version of Zignal’s software utilizes AI to ‘scour global digital data,’ a phrase that suggests a preference for avoiding the term’ mass data collection.’ Two months later, ICE signed the contract.

If you connect the dots, it looks less like a one-off purchase and more like a continuing build-out of a federal AI surveillance infrastructure – a system built by private companies, financed by government budgets, and justified by the language of ‘threat detection.’ 

Politics and Pattern Recognition

The timing matters. 

Under Trump’s administration, ICE has grown bolder in linking immigration enforcement to online behavior. Pro-Palestinian activists like Mahmoud Khalil were detained after being doxed on right-wing sites such as Canary Mission. 

More recently, ICE raids in New York followed a viral post from a right-wing influencer demanding a crackdown on street vendors.

What’s changing now isn’t just who ICE targets, but how it identifies them. When AI begins labeling ‘risk’ based on social media chatter, political speech becomes data, and data becomes a potential trigger for enforcement.

Civil rights groups have already pushed back. A coalition of labor unions and the Electronic Frontier Foundation recently sued the federal government over what they call ‘viewpoint-driven surveillance.’ 

The lawsuit argues that AI monitoring chills free expression by making people think twice before posting about controversial topics – or, more accurately, before posting anything at all. 

The Tech Behind the Curtain

Zignal’s platform is a big-data engine powered by machine learning models that scrape, classify, and rank billions of posts from Twitter, Facebook, YouTube, Telegram, TikTok, and obscure corners of the internet you may have never heard of.

Each post gets analyzed for keywords, geolocation clues, network connections, and ‘narrative trends.’ 

Then the system generates automated alerts – the ‘curated detection feeds’ ICE will now receive. The problem is that these models aren’t trained to handle nuance. They flag ‘signals,’ not context.

If an algorithm sees a spike in a hashtag related to Gaza protests, it can tag that as ‘emerging unrest.’ A cluster of accounts talking about migrant rights might be labeled a ‘coordinated network.’ 

What happens next depends on who’s reading the dashboard, and how eager they are to show results. 

As Patrick Toomey from the American Civil Liberties Union’s National Security Project put it, ‘The Department of Homeland Security should not be buying surveillance tools that scrape our social media posts and use AI to scrutinize our speech.’ 

But that’s exactly what’s happening. And it’s being done quietly, without public oversight or disclosure of what’s being monitored. 

From Social Media to Social Control

Every technology wants to scale. Surveillance tech, especially so. Once the system is in place, the temptation to use it more broadly is irresistible.

ICE isn’t the only agency expanding its AI footprint. 

In the same week as the Zignal deal, ICE signed a $7M contract with SOS International (SOSi) for ‘skip tracing.’ Basically, tracking people’s whereabouts using digital footprints. 

Two months earlier, SOSi had conveniently hired ICE’s former intelligence chief, Andre Watson, to help ‘deliver capabilities’ to law enforcement clients.

It’s a revolving door made of machine learning and public contracts. 

The same people who design the government’s surveillance playbook end up selling it back to the government for millions.

Meanwhile, the AI models that power these systems are opaque, prone to bias, and nearly impossible to audit. The more data they ingest, the more confident they appear, even when they’re wrong. A misplaced flag or an overzealous analyst can turn a tweet into probable cause.

And yet, politically, AI surveillance remains one of those bipartisan comfort zones. Democrats call it modernization. Republicans call it law and order. 

Everyone calls it ‘data-driven decision-making.’ Few call it what it is: automated suspicion.  

Why This Matters for Tech Policy

The ICE-Zignal deal is a case study in how fast the surveillance market is merging with the AI industry. Five years ago, ‘AI-driven social monitoring’ sounded like marketing jargon. Now it’s a procurement line item.

For tech policy, the implications are huge. The government’s appetite for predictive intelligence means there’s steady demand for companies willing to turn the internet into an open-source intelligence feed. 

That’s lucrative for Silicon Valley firms that once sold brand sentiment analysis – they just rebrand it as ‘national security analytics.’

The losers are privacy, transparency, and democratic accountability. 

When an algorithm decides which posts are ‘risks,’ the targets have no way to appeal, correct, or even know they’ve been flagged. The datasets are proprietary, the models are secret, and the public has no seat at the table. 

When Surveillance Becomes Routine

The ICE contract doesn’t just show how government surveillance evolves. It shows how it normalizes. 

AI makes monitoring feel efficient, clean, and automated, stripping away the human decision-making that once made surveillance controversial.

Once a tool like Zignal Labs is embedded in federal systems, it becomes difficult to remove. Agencies get addicted to the data flow, politicians point to ‘threat dashboards’ as proof of vigilance, and taxpayers foot the bill.

The border between public safety and political policing is becoming increasingly blurred through the use of algorithms. For a system that can analyze eight billion posts a day, it’s ironic how little it seems to understand about human rights.