Elon Musk’s DOGE team seems to be rifling through the federal bureaucracy at breakneck speed, and inside sources say that is attributable to the use of AI software. These anonymous sources, speaking to Washington Post reporters, also claim that federal data is being fed into AI systems haphazardly and in a way that might create serious security and privacy risks.
The claim comes from sources at the Department of Education (DOE), which seems to be a particular point of focus for Trump and the DOGE team. While the agency cannot be abolished without an act of Congress, Trump appears to be set on gutting it to as great a degree as possible with executive orders. That has translated into targeting the payroll and the agency’s various programs for very deep cuts, which DOGE has apparently relied on unspecified AI software to identify. Whether actual lawbreaking has taken place remains unclear, but the incident has once again put the spotlight on AI in the workplace and the risks of feeding sensitive internal information to it.
How secure is DOGE’s use of AI software?
It remains unclear if DOGE has committed any violations of law or privacy regulations in its handling of federal data, as the anonymous sources did not name specifically what AI software is being used or how it is being used. The concern is understandable, however, after over two years now of workplace incidents involving employees packing sensitive company data off to ChatGPT and other similar LLM services.
If federal data is being sent to third party AI software in this way, it could find its way into training data. That means it could appear somewhere else, either at random or as the result of hackers figuring out how to slip the system’s guardrails. Numerous companies, particularly those in the financial, medical and legal industries, have banned workplace use of LLMs for exactly this reason.
AI software is a powerful tool for these tasks if used responsibly; the big question is if the DOGE team is staying within the bounds of the law. The types of federal data they are necessarily encountering often have special protections, and the audit team may well be facing its own audit before long.
DOGE team using Azure to organize efforts, but full use of AI in scanning federal data remains unclear
The inside sources have said that DOGE is using Microsoft Azure to centralize their research and scanning efforts, but the platform allows for access to a wide variety of types of AI software. The team has reportedly ordered “lower level staffers” to furnish it with access to financial information at the DOE, but it is unclear if that information has been run through some sort of third-party AI.
While DOGE has made the Treasury and DOE two of its biggest points of early focus, Trump has promised that this scanning model will be going through most of the rest of the federal government at some point and that government contracts will definitely be subject to some sort of AI software scanning. The DOE item that seems to have critics most concerned among the collection of federal data is the student loan database, which contains tens of millions of Social Security numbers and payment information paired with contact information for current and former borrowers.
While opinions differ on DOGE’s drastic overhaul of the federal budget, Americans should have a reasonable assurance that their personal data is protected during the process. It remains unclear what best practices, if any, the team is following in terms of data protection.
