100 Days of DOGE: Assessing Its Use of Data and AI to Reshape Government

Shortly after the November 2024 United States election, then President-elect Donald Trump announced that his billionaire patron, Elon Musk, would lead a new “Department of Government Efficiency (DOGE)” intended to "dismantle Government Bureaucracy, slash excess regulations, cut wasteful expenditures, and restructure Federal Agencies—essential to the 'Save America' Movement." On his first day in office, President Trump established DOGE by issuing an executive order that repurposed the US Digital Service “to implement the President’s DOGE Agenda, by modernizing Federal technology and software to maximize governmental efficiency and productivity.”

100 days into the Trump administration, DOGE has emerged as one of its most consequential and controversial initiatives. While it has failed to achieve its stated goals around cost savings, it has succeeded in reducing the size of the federal workforce by tens of thousands of workers, gutting entire agencies along the way. There are substantial questions about the constitutionality of DOGE, and it is the subject of multiple lawsuits that challenge its operations. There are also concerns over corruption: a Senate Permanent Subcommittee on Investigations minority staff report suggests Musk’s cuts to the government agencies that regulate his companies may save him and his businesses more than $2 billion in potential liabilities, though that figure “drastically understates the true benefit Mr. Musk may gain from legal risk avoidance alone as a result of his position in government.”

Central to Musk’s effort has been DOGE’s commitment to integrating AI into its investigations and government systems.

Since DOGE’s official launch in January, the group has leveraged AI in two primary ways: utilizing the technology to analyze government data and developing internal tools for federal agencies. A key goal is to automate as many government operations as possible. As The Washington Post reported in February:

“The end goal is replacing the human workforce with machines,” said a US official closely watching DOGE activity. “Everything that can be machine-automated will be. And the technocrats will replace the bureaucrats.”

Below is a summary of key details about DOGE, including its use of AI and handling of sensitive government data.

Combining and analyzing federal data

In the first 100 days of the Trump administration, DOGE has had unprecedented access to federal datasets and citizen information. This access has been the subject of widespread scrutiny and multiple lawsuits, primarily on the basis that DOGE’s access violates the 1974 Privacy Act, a post-Watergate statute that prohibits federal organizations from sharing “sensitive information with other parties, even within the federal government.” While the majority of these lawsuits are still somewhere in the legal process Cases that implicate DOGE, such as the Electronic Frontier Foundation’s suit against the Office of Personnel Management (OPM), DOGE, and Musk that challenges DOGE’s access to OPM information, are still ongoing, and would help set precedent for the scope of the 1974 Privacy Act and the legal limits for DOGE’s access to data.

While it remains unclear exactly how DOGE is utilizing federal data for AI models, pro-democracy groups such as Democracy Forward have submitted Freedom of Information Act (FOIA) requests regarding the “reported use of AI by DOGE and the Trump Administration,” according to CNN. In a statement on the FOIA requests, Democracy Forward CEO Skye Perryman stated that, “The American people deserve to know what is going on – including if and how artificial intelligence is being used to reshape the departments and agencies people rely on daily.” The ACLU filed a similar request targeting documents related to DOGE's access to data at the Department of Veterans Affairs.

Surveillance of government employees

Worryingly, DOGE has reportedly used AI tools to surveil the conduct of federal employees, particularly for behavior or work that may contradict President Trump’s agenda. As of early April, officials in the Trump administration “told some US government employees that Elon Musk's DOGE team of technologists is using artificial intelligence to surveil at least one federal agency’s communications for hostility to President Donald Trump and his agenda,” according to Reuters. On April 10, the Guardian reported that organizations that work with the Environmental Protection Agency “sent a warning to staffers that their meetings and phone calls with the agency were being monitored by an artificial intelligence tool.” The EPA has since confirmed that it has been “looking at AI to better optimize agency functions and administrative efficiencies” but denied using AI tools “as it makes personnel decisions in concert with DOGE.”

Targeting immigrants

According to CNN, DOGE is developing a comprehensive database to expedite immigration enforcement and deportations by integrating sensitive data from multiple federal agencies, including the IRS, the Social Security Administration, and the Health and Human Services (HHS). Wired reported on the databases DOGE gained access to at HHS. The project, supported by the Silicon Valley company Palantir, aims to extend its scope beyond criminal investigations to identify individuals with civil immigration violations, thereby significantly broadening its data-driven approach to immigration control. (There is already some evidence of DOGE “assisting” in the investigation of immigrants.)

Deploying AI tools in government agencies

DOGE has been clear about its intention to roll out artificial intelligence across the federal government. Thomas Shedd, a former Tesla employee who was initially appointed by the new administration to head technology at the General Services Administration and is now the Chief Information Officer at the Department of Labor, told GSA staff that “AI would be a key part of their cost reduction work,” according to the New York Times. However, it has been less clear whether and how AI has been rolled out across federal government agencies. Below are some of the reported AI uses by DOGE in the first 100 days of the Trump Administration:

• CamoGPT: This AI tool, reported by Wired, is said to be used by the Army to scan its records systems for any references to DEIA programs. The Army confirmed in an email to Wired that the tool was being used, but did not provide further information about how the model works or its use cases.
• GSAi: The GSA launched a chatbot, “GSAi,” to 1500 employees at the agency. It has been internally framed as a “productivity booster” to fill the gaps left by fired employees. Thomas Shedd stated that “as we decrease [the] overall size of the federal government, as you all know, there’s still a ton of programs that need to exist, which is a huge opportunity for technology and automation to come in full force,” according to The Atlantic. Shedd suggested “coding agents,” such as GSAi, have the potential to be used across the government to replace employees. Shedd also stated that GSAi could be “run on contract (and) automate GSA’s finance functions.” According to a GSA spokesperson, GSAi has been under development over the past 18 months under the Biden Administration, and has been developed “in-house” to help mitigate trust and safety concerns.
• Possible LLM usage to assess the “5 Bullet Points” emails: On February 22, OPM sent out an email to all federal employees demanding that they respond with “5 bullet points listing what they accomplished this week,” at risk of losing their job. Three sources with “knowledge of the system” reported to NBC that the responses are fed into a Large Language Model to determine “whether someone’s work is mission critical or not.” Elon Musk has denied these claims on X, saying that LLMs were “not needed here” and that the emails were to “check to see if employees had a pulse.”
• Use of AutoRIF: DOGE is said to be accessing and editing the AI tool AutoRIF (Automatic Reduction in Force) to “assist in mass firing of federal workers.” AutoRIF was developed by the Department of Defense over 20 years ago and has been used by a variety of agencies to assist in workforce reduction efforts in the past. However, screenshots of internal databases reviewed by Wired appear to demonstrate that AutoRIF’s source code is being edited by DOGE.
• Department of Education records: DOGE has been feeding “sensitive data from across the Department of Education into artificial intelligence software” to analyze programs and spending within the agency, reported the Washington Post. According to the report, the software is hosted on Microsoft’s Azure and is being used to review all the funds dispersed by the Department.

Going forward

Journalists, lawmakers, civil society groups, and other watchdogs are seeking to obtain more information from DOGE and its interventions across federal agencies. In the weeks ahead, expect more details to emerge about DOGE's efforts to access and combine government data for various purposes, as well as to deploy AI systems within government agencies, even as it continues to dismantle them.

For instance, early last week, 48 Democratic lawmakers signed a letter addressed to Russell Vought, Director of the Office of Personnel Management, demanding information about DOGE’s “unauthorized use of artificial intelligence systems” in the federal government. The letter cited the use of private data in generative AI models at agencies such as OPM and the Department of Education, warning that, “without proper protections, feeding sensitive data into an AI system puts it into the possession of a system’s operator—a massive breach of public and employee trust and an increase in cybersecurity risks surrounding that data.” The letter also contained and requested answers to 10 questions, most notably asking DOGE and the Trump Adminsitration to specify exactly how they are using AI (including any use of Musk owned GrokAI), who has access to those models and their data, as well as the use of if AI has had any role in DOGE’s cost cutting efforts. At this time, there has been no acknowledgment of the letter by the Trump Administration or DOGE.

Even as DOGE continues to target a wide range of components of the federal government, from the Postal Service to the Department of Defense, Elon Musk has announced he will reduce his effort on the project to focus on his businesses, including Tesla, which has been the target of sustained protests that contributed to billions in losses to its market value. But while it is too early to assess the long-term impact of Musk’s initiative, it appears clear to many observers that DOGE and its use of AI are fundamentally anti-democratic. As Tech Policy Press fellow Eryk Salvaggio wrote in February, DOGE is operating on the false premise that “the goal of bureaucracy is merely what it produces (services, information, governance) and can be isolated from the process through which democracy achieves those ends: debate, deliberation, and consensus.” Restoring democracy in the US may eventually mean reversing much of what DOGE has wrought.