That may be a legally ordinary answer. It is also the democratic problem in miniature. Once the operating layer is installed, the issue is no longer only what the vendor intended. It is what the institution can do faster than oversight can follow.
The military side is larger and less visible. In July 2025, the Army announced a Palantir enterprise agreement to support future software and data needs. Reuters reported that the agreement could allow up to $10 billion in purchases over ten years, though the Army was not required to spend the full amount.⁸ Defense Department records also show a $795 million modification for Palantir’s Maven Smart System work, extending into 2029.⁹
Those figures do not prove conspiracy. They show something more ordinary and more durable: dependence. Dependence arrives through budgets, renewals, training, habits, and the simple fact that once an agency learns to see through a system, it becomes difficult to imagine seeing without it.
There is precedent. IBM’s punch-card systems did not create the modern bureaucratic state, or the crimes later committed by states that classified people. They did show how administrative power changes when government can sort populations faster and at larger scale. After September 11, fusion centers and watch-list systems offered another lesson: once agencies are told to connect dots, data sharing can outrun the discipline required to govern it.
The old paper problem was fragmentation. The new digital problem is action without enough institutional drag.
That is why Palantir is more than routine outsourcing. Lockheed builds aircraft. Raytheon builds missiles. Consulting firms redesign agencies and leave binders behind. Palantir operates closer to the decision surface, where records are joined, patterns displayed, options ranked, and work routed to the next official. It does not merely supply a tool used after policy is made. It helps shape the field in which policy becomes possible.
The IRS controversy pushed the issue into civilian life. In June 2025, Senator Ron Wyden, Representative Alexandria Ocasio-Cortez, and other lawmakers wrote to Palantir after reports that its employees were connected to work involving a “single, searchable database” or “mega API” for taxpayer information.¹⁰ Tax records contain work, family, income, dependents, addresses, losses, debts, and the shape of a private life. Once those records become easier to join with other government systems, government gains capacity before the public has seen the architecture.
Palantir denies the darkest version of that story. The company has said it is not building a “master list” of Americans and has argued that its government work is lawful, bounded by agency authority, and constrained by privacy rules.¹¹ That denial belongs in the record.
But the strongest concern does not require proof of a completed master database. It requires attention to the visible pattern: multiple integrations, growing contracts, executive pressure to eliminate data silos, and a company positioned where speed meets state power.
The defenders of this architecture have a serious argument. Government failure is not neutral. Obsolete systems lose benefits claims, miss fraud, slow military response, obscure criminal networks, and bury warnings in incompatible databases. A democratic state that cannot see what is happening inside its own records becomes weak in ways that hurt ordinary people first.