During the rapidly unfolding constitutional crisis of the past three months, nearly every day has brought a new, alarming development. While many of these actions — tariffs imposed on uninhabited islands, the “mistaken” letter of demands sent to Harvard, Hegseth’s texting habits — reflect an equal combination of incompetence and malice, two of the most worrying developments appear more strategic and potentially despotic: DOGE’s takeover of the federal bureaucracy, and the buildout of new state machinery to enact Trump’s goals.
What these policies have in common, and what the actors carrying them out understand better than most, is that all governance is now data governance. To understand the authoritarian threat presented by Trump 2.0, we need to understand how the Silicon Valley contingent of his administration prioritizes and uses informational power. For whatever their limitations and ignorance about how agencies operate, Elon Musk and the DOGE minions understand the importance of control over data flows to modern governance. Indeed, though DOGE has been publicly pitched as an effort at achieving monetary efficiency (a laughable claim as they’re saving only a fraction of what they originally promised and potentially costing much more), this was never their true goal.
Instead, DOGE is hard at work producing (or attempting to produce) another kind of efficiency — the centralized information architecture needed for DOGE’s worker-free, AI-first vision of public administration. To be clear, an efficient tool can still be used in laughable ways. Applying bad AI to your data will produce stupidity — for example, DOGE’s automated scrubbers, likely set to erase references to LGBT history, recently removed the page on the Enola Gay bomber that dropped the first atomic bomb.
But such idiocy should not distract us from the significance of the goal. Setting dubious claims of AI and mass automation aside, highly integrated information systems can still be extremely powerful mediums of governance. Technical infrastructure can provide astonishingly efficient means for enacting an agenda. They remove the barriers of logistics, lowering the costs of time and coordination and reducing opportunities for checks and accountability.
DOGE understands that he who controls the data infrastructure holds the keys to the government. Want to stop, or even claw back a federal payment you don’t like? You can do that. Exclude undocumented immigrants from financial life? Easy. Want to track union opposition, and obtain sensitive information on your competition? Done. Rig elections? A helpful step would be to compel access to state-held voter rolls. Need to find all immigrants? They are working on that too.
Consider this last notable example. Immigration law has long extended extraordinary powers to executive authority, and Courts remain broadly deferential to how the executive exercises that authority. The biggest check on mass deportations is not the law — it’s the logistical hurdle of locating documented and undocumented migrants in the first place. You can’t deport those you can’t find.
This makes recent news that Palantir is working to consolidate federal data and build a centralized database to locate and track immigrant populations all the more alarming. With a centralized database, Palantir can locate — perhaps even in real time — immigrant targets. Similarly alarming is the announcement that tax data from the IRS will be shared with ICE — an agreement that flouts longstanding IRS practice against sharing taxpayer information with other government agencies, and which caused a senior IRS official to resign in protest.
Data flows can facilitate enforcement goals by removing the barriers of enforcement — locating a target, confirming their identity, and coordinating across federal and local enforcement agencies. But informational power can go further. It can also be used to discipline those disfavored by the administration by toggling how information mediates access to benefits.
By placing the names of immigrants on the Social Security Administration’s “official death” file, the Trump Administration can immediately cut people off from crucial financial services, including banking, credit cards, and access to state benefits. Weaponizing information systems designed to facilitate benefits is doubly harmful, as it leaves immigrant residents in a bind: state legibility is necessary to participate in commercial and civic life, while also carrying with it risks of state violence. In fact, it’s because information is such a powerful bureaucratic tool that our modern federal privacy laws were enacted to limit and restrain its exercise. For instance, when the Johnson administration proposed consolidating hundreds of federal databases into a centralized “National Data Bank” in 1965, public backlash eventually led Congress to adopt the Federal Privacy Act of 1974, which forbids agencies from disclosing individuals’ records across agencies without prior written consent.
Those privacy laws are, for now, restricting full DOGE access to Social Security systems — systems which contain banking information, school records, salary and mental health records for nearly everyone in the country. But in other instances, DOGE has taken a “break things first, apologize later” approach to both the privacy laws protecting federal information systems and the systems themselves. At Treasury—and despite a government lawyer’s insistence in court to the contrary—DOGE operatives were granted direct access to the federal payments system.
As DOGE trammels informational protections, it not only gains access to informational power to enact Trump’s agenda, but also undermines the conditions for future information sharing on which good governance depends. Take, for example, the disclosure of personal information of undocumented immigrants paying into the tax system. Trust is required for people to share information, and sharing information is inextricable from the underlying function of the tax system — namely, collecting taxes. Hence the strong privacy protections traditionally afforded taxpayer information. DOGE’s thuggish disregard for those protections is as much a threat to the IRS’s ongoing functionality as the massive reduction in the IRS’s enforcement staff.
The Silicon Valley authoritarians in the Trump regime know this lesson in the reverse too. The same Trump administration that is concentrating and combining sensitive data on federal workers and citizens is halting all manner of public agency data collection — gutting climate data collection at EPA and the NWS, eliminating data on substance use, mental health, and the rising maternal mortality, and dismantling expert panels in order to manipulate data on the economy. Techno-authoritarians know that it is not enough for them to amass their own informational power. To achieve their goals, they must also sabotage the informational basis for accountability and for administrative expertise and capacity.
By eliminating teams that create and share valuable forms of federal data collection, DOGE is clearing the path for for-profit private alternatives (a longstanding right-wing dream for the National Weather Service), while hampering our public capacity to measure and understand American society (e.g. tracking maternal mortality in a post-Dobbs America). These attacks on federal data also happen alongside a broader attack on the creation and dissemination of knowledge for the public good, embodied in the recent attempt to gut grants for library services, and the dismantling of the federal system of basic scientific research support.
So far, the left has mostly been playing defense against this powerful vector of governance in the courts, attempting to stop DOGE from amassing the breadth and depth of informational power that Musk hopes to control. While defense is absolutely the correct posture in times like these, it’s also worth asking whether our overly cautious approach to positive governing capacity — in information, as elsewhere — has achieved much, apart from giving us yet another rule over which to cry foul. This time of crisis, as Luke Herrine has urged, is also a time for lefty researchers to decide what systems we want to build out of the current wreckage, and what vision of data governance and informational power they require to enact.
