A United States Army general speaking with a contractor at a military trade show in 2023.
In June 2015, the United States government announced what remains one of the most catastrophic data breaches in American history. The agency responsible for managing the federal workforce and processing security clearances, the Office of Personnel Management, was breached by hackers. Newspaper reports afterwards linked the breach to Chinese intelligence. The stolen data included security clearance files and SF-86 forms for millions of federal workers. The SF-86 form, by the way, is an exhaustive questionnaire that documents the most intimate details of a person's life to include foreign contacts, mental health history, financial struggles, past drug use, and family vulnerabilities. And even biometric data such as fingerprint records were part of the breach. In intelligence terms, this was not merely a data breach. It gave a power with little regard for individual rights and one of America's most committed adversaries a playbook for compromising millions of America's most trusted workers. Foreign intelligence services now possessed the tools to identify, pressure, blackmail, or expose American intelligence officers, diplomats, and military personnel worldwide. The damage is permanent. Eight years later, in early 2023, roughly three terabytes of internal US military emails, many connected to US Special Operations, were gathered by entities unknown and unauthorized to have them. The reported access point was a server hosted on Microsoft's Azure government cloud. Although it was physically separated from commercial infrastructure and cleared for military data, it was later discovered to have been sitting open on the public internet without a password for two weeks. Much of the public reaction for both events focused on government incompetence. But that framing misses the more important story. In neither case was the failure primarily one of government sloppiness. In the OPM breach, the data did live on government-controlled infrastructure. However, the failure came through the fact that the keys to that infrastructure were held by private contractors. And it was through those contractors that foreign intelligence walked in and took it.
The relationship between the US government and private vendors is not new. During the Revolutionary War, the Continental Army relied on private merchants for weapons, food, and logistics. World War II was won in significant part by the industrial capacity of private American manufacturers such as Ford, Boeing, and Grumman. Those companies converted their facilities to wartime production at the government's direction. But in every era, the lines were clear. Private companies supplied materials and capacity. Government retained authority, accountability, and control. That began to change in the 1990s, when a wave of outsourcing ideology swept through Washington under the banner of efficiency. Functions once considered inherently governmental from prison management to military logistics were handed to private firms. Then came September 11, 2001, and the contracting crossed the line into outsourcing and never came back. The wars in Afghanistan and Iraq introduced something unprecedented almost every other week. One of the biggest moves was the shift to private military contractors, most notoriously Blackwater, that were given the authority to use lethal force with no accountability. Blackwater operatives operated in war zones, killed civilians, and faced neither military justice nor meaningful civilian prosecution for years. The normal chain of command, the rules of engagement, and the accountability that defines military service were not applicable to the private companies. That same post-9/11 period saw the government outsource the background investigation process itself. The work of determining who is trustworthy enough to access America's most sensitive secrets. This was always considered a core sovereign function. But it was handed to private firms like USIS and KeyPoint Government Solutions. These contractors conducted the vetting.
Representative Chris Stewart of the House Permanent Select Committee on Intelligence stated there was direct evidence China was cross-referencing the personnel and security clearance files stolen from the Office of Personnel Management in 2015 for intelligence purposes.
For most of American history, background investigations for government security clearances were conducted by the Federal Bureau of Investigation (FBI). Agents interviewed neighbors, former employers, family members, and associates. The files they compiled never left government hands. Every step of the process, from investigation to adjudication to records storage, was handled by federal employees operating under clear legal authorities and oversight. The same philosophy governed government information systems. When agencies purchased computers or software, they owned what they bought. Government technicians knew every line of source code, every command, and every architectural decision. There were no black boxes. Vendors supplied the hardware and tools while government personnel operated them. What has replaced that is something unrecognizable. Not only are private contractors now conducting the investigations and holding the access credentials, but just as disturbingly, the databases and systems themselves are increasingly operated by vendors whose internal architecture the government does not fully control or even comprehend.
That problem has now extended into artificial intelligence. The Department of Veterans Affairs has integrated AI tools into claims processing, making algorithmic recommendations that affect the benefits of millions of veterans. The Social Security Administration has deployed AI in fraud detection systems that can flag, delay, or deny assistance to citizens. The Department of Homeland Security uses AI-assisted tools in immigration enforcement and threat assessment. In each case, the underlying system is a commercial product, developed by a private company, plugged into sensitive government databases. The penetration of AI into defense and intelligence operations has moved far beyond administrative functions. The country's major AI firms including OpenAI, Google, and Amazon Web Services have signed deals with the Pentagon to deploy their AI systems directly into the Department of Defense's classified networks. The Army itself warned in 2025 that a battlefield communications system designed by Palantir and Anduril was a "black box" that makes it impossible to determine whether unauthorized users can access its applications. But even amid warnings and the startling statements from CEO's about not fully understanding their own product, the US Government continues plugging them into America's databases. For most of American history, no decision affecting a citizen's rights, benefits, or freedom was delegated to a system the government did not build or control.
The AI systems now being integrated into government operations are not simple rule-based programs whose logic can be audited line by line. They are large neural networks whose decision-making processes are opaque even to their creators. We cannot fully audit their reasoning. And yet these are the systems being connected to federal systems that inform decisions on everything from war to tax returns. The outsourcing of American government began with supply contracts and grew into something that looks more like another arm of government itself. We have now reached the point where the most sensitive functions of the state are controlled by systems that private entities control with the government more and more simply acting as a contract manager.
If you find any of the articles useful, please share with others. Also, send an email so we can keep in touch for special articles and episodes. Email: jeremy@plainmeaning.com
