In the Shadow of Power - How Trump and Thiel Are Building a Digital System of Control - An Investigative Report

The new security state has no pomp. It stands behind manicured hedges, wears glass facades without a nameplate, and works with data we produce voluntarily. In Vienna, Virginia, just a few miles from the CIA, the Threat Screening Center (TSC) maintains the national watchlist - a legacy of the post-9/11 era, now rebuilt for the present. And in an unassuming business park in Williston, Vermont, Immigration and Customs Enforcement (ICE) is setting up a social media hub designed to turn posts, geotags, and friend networks into actionable leads. Together, these locations outline the contours of a system that no longer deals with terrorists of the old kind but redefines the boundaries between internal security, political persecution, and everyday digital surveillance.

The center in Virginia is the heart of it. It manages the federal watchlist, whose criteria and logic of inclusion remain opaque. Even the director’s name was not previously public; internally, he is known as Steven McQueen, a longtime FBI counterterrorism expert. From the White House comes the new rhythm: National Security Presidential Memorandum 7 (NSPM-7) directs the security apparatus to monitor “indicators of violence” - categories that include “anti-Christianity,” “anti-capitalism,” and “anti-Americanism.” The reinterpretation is programmatic: what was once political opinion protected by the Constitution is now turned into an early warning signal and fed into a database.

While ICE and the Threat Screening Center officially speak of “modernized analytical platforms,” it is already clear who provides the technological backbone: Palantir Technologies, the data company of Trump confidant Peter Thiel. The contracts with ICE show that the infrastructure is built on three core software components - Gotham, Foundry, and Metropolis (formerly known internally as “Golem”). Gotham handles operational case analysis, Foundry integrates and evaluates large datasets, and Metropolis performs algorithmic pattern recognition. Together, they form a system that enables authorities to construct complex personal profiles and risk indicators from scattered fragments of information.

What makes it explosive is that Thiel’s technology has already taken hold in Germany - not with intelligence services, but with police authorities. Several federal states, including Hesse, North Rhine-Westphalia, and Bavaria, use Palantir’s Gotham platform under the project name HessenDATA or similar variants for data analysis. Baden-Württemberg also plans to join: the state government has already signed a framework agreement with Palantir to deploy the software in its police force - subject to legal adjustments currently being prepared. While data protection advocates and civil rights groups warn against the expansion of preventive surveillance, state authorities emphasize efficiency and security. Yet the direction is clear: what is promoted in the United States as a tool against “national threats” is becoming in Germany part of a new, data-driven policing strategy - and shows how seamlessly surveillance logics spread across continents.

The numbers illustrate the shift. When Donald Trump returned to office, the watchlist contained around 1.1 million entries, mostly foreigners, less than half a percent of them “U.S. persons.” A senior intelligence official we spoke to said that the idea of a genuinely domestic watchlist had long been blocked by the rule of law: Americans could not be comprehensively monitored or placed on a list without a legal predicate. The lessons from Watergate, the Church Committee, FISA reforms, and the Snowden years had created a web of rules separating political beliefs from criminal acts. “Until now,” he said. The expansions introduced under Obama and Biden - such as the inclusion of transnational organized crime - laid the foundation. NSPM-7 provides the political frame to turn it into an omnipresent grid.

In parallel, the label was adjusted. The old Terrorist Screening Center was already renamed the Threat Screening Center at the end of Trump’s first term - a semantic shift that turned the fight against terror into a fight against “threats” in general. After the change of administration, the name was quietly reverted, and now, since March, officially expanded again. FBI Director Kash Patel called it a signal: the list would grow larger, its scope broader, and cooperation with agencies at all levels tighter. After the murder of Charlie Kirk, Patel boasted before Congress of a 300% increase in domestic terrorism investigations; a large portion now falls under “Nihilistic Violent Extremism,” a stretchable category that can contain as much as the executive chooses to put in it.

While Vienna builds the categories and pulls the threads, Williston provides the raw material. ICE plans - alongside a sister site in Santa Ana, California - to hire at least a dozen external contractors to systematically evaluate open web sources and social media. Facebook, Instagram, X, along with commercial data pools, police, and government databases, are to be fused into profiles that field agents can use: locations, movement patterns, associated persons. The drafts, published as a “Request for Information,” set the course: until now, the yield from open sources had been “limited”; now the agency wants to professionalize it - with people who are not on payroll but under contract. Start date: as early as May 2026. In Williston, the Law Enforcement Support Center already operates, a nationwide hub for information requests and a major tip line; a few miles north, in St. Albans, is the field office where reporting requirements are enforced and, in some cases, overnight detentions have been ordered. Vermont is no outpost - it is both laboratory and node.

The logic of the division of labor is striking: what ICE distills as a “lead” from posts, geotags, and contact lists flows into the national grids - and is compared against categories that the TSC is currently redefining. Below, in the local Joint Terrorism Task Forces, categories become case files; above, in Vienna, files become classes of threats. Between them stand Palantir tools, newly commissioned this year, to “optimize” the pipeline from identification to prioritization to operational action.

With masterful coldness, this leaked screenshot shows what Washington’s architecture of power protects in an emergency: not administration, but control. It is a sober administrative document - and at the same time a mirror of the political reality of a country that leaves its security apparatus untouched even in a shutdown. The highlighted line shows the FBI, whose numbers stand out: 36,755 employees, and all 36,755 are classified as “excepted” - fully exempt from suspension. This means that even if the rest of the federal government shuts down, the FBI remains fully operational. The screenshot breaks this down further: over 13,000 special agents, more than 3,000 intelligence analysts, over 200 attorneys, and over 20,000 additional specialists would continue to work. All divisions of the Department of Justice are listed - from the Drug Enforcement Administration (DEA) to the Bureau of Prisons (BOP) to the FBI - showing for each how many employees they have in total (“Total On Board”) and how many would continue working during a shutdown (“Excepted Employees”).

Legally, the system is built on a double foundation. The watchlist is not a criminal register; it asserts no proof of guilt and yet carries consequences: additional screenings, travel bans, secondary inspections, secret designations. Because the criteria are secret, there is no path to effective appeal. Because private contractors curate the raw data, the line where constitutional boundaries apply becomes blurred. And because “threat” is not a clearly defined term, politics becomes the grammar of risk. Those who decide what is “nihilistic” also decide who is to be controlled.

That is where the democratic imbalance begins. Stephen Miller, Trump’s security adviser, has already made the direction public: he speaks of a “whole-of-government” effort against “left-wing terrorism” - a framing that abstracts from the real phenomenon of leftist violence to turn protests, sit-ins, or ICE blockades into parts of a “terror ecosystem.” It is the semantic groundwork that makes data categories plausible. And it is the political sleight of hand that turns the appearance of legality into an administrative practice whose effectiveness lies precisely in its invisibility.

Vermont reacts as free societies should: with skepticism. The ACLU of the state warns that the expansion of digital surveillance in Williston means large parts of the population are being swept into the net - not because of specific suspicion, but because their digital existence leaves traces. This is no alarmist exaggeration but the functional principle of open-source intelligence: the more ordinary the sharing of locations, images, and relationships becomes, the easier everyday communication can be translated into operational hypotheses - and the greater the temptation to turn probability into reality.

The Trump administration sells this as a response to a “new threat landscape.” But those who look closely see less a new reality than a new sequence: first political categories are defined, then the technology is adapted, and finally the demand for data arises. The sequence is crucial because it explains why the state tends to make ever greater use of its own tools. Those who have a watchlist will fill it. Those who build a social media lab will feed it. And those who define “threat” more broadly than “terror” will turn more people into data points than a democracy can bear.

This system has no center of steel and concrete that one could visit. It consists of interfaces, contracts, forms, and guidelines. In Vienna, behind a black fence, the grammar is written. In Williston, between gas stations and warehouses, the sentences are filled with content. Between them, millions of data points flow through Palantir’s logic trees, while down the street, the Joint Terrorism Task Forces execute the operational implementation. It is a network that does not stand out for its hardness but for its elasticity - and that is why it is so hard to grasp.

One can defend this architecture of control by invoking real dangers. One can criticize it by invoking the history of American civil rights. To understand it, one must do both - and then ask whether the state still provides the proof it has owed its citizens for decades: that it does not turn the power it holds against its critics. The new watchlist logic taking shape simultaneously in Virginia and Vermont reverses the burden of proof. It no longer asks what someone has done, but what their data might suggest. That is what makes it so effective. And that is what makes it so dangerous.