House Democrats are demanding more information about how artificial intelligence may be used inside federal agencies as DOGE pushes automation during major workforce cuts. Their central concern is direct: sensitive government data, including health, financial, biographical, and student aid information, could be exposed to AI systems without clear public accountability.
The requests, first obtained by WIRED, were sent Wednesday morning by Democrats on the House Oversight Committee. They were signed by Gerald Connolly, a Democratic congressman from Virginia, and directed at federal agency leaders.
What Democrats Want Agencies to Explain
The committee Democrats sent two dozen requests asking agencies to describe plans to install AI software across the federal government. The inquiries follow reporting by WIRED and The Washington Post about efforts by Elon Musk’s so-called Department of Government Efficiency, known as DOGE, to automate work with proprietary AI tools and gain access to sensitive data.
The requests ask agencies to show that any AI use is legal and that protections are in place for private information. The concern is not only whether AI tools are being deployed, but whether agencies can explain what the tools do, what data they touch, and what safeguards apply.
The requests state that the American people give the federal government sensitive personal information related to “their health, finances, and other biographical information” with the expectation that it will not be disclosed or misused without consent, including through “an unapproved and unaccountable third-party AI software.”
That framing puts data protection at the center of the dispute. Federal agencies hold information that can be deeply personal, and the Democrats are asking whether AI systems connected to DOGE are being introduced before agencies have answered basic questions about approval, oversight, and security.
The Musk Conflict Question
The requests also raise a separate issue: whether the use of AI inside federal agencies could financially benefit Elon Musk. The source article notes that Musk founded xAI and that Tesla, described as his troubled electric car company, is working to pivot toward robotics and AI.
Connolly’s concern is that Musk could use access to sensitive government data for personal enrichment. The requests specifically point to the possibility that data could be used to “supercharge” Musk’s proprietary AI model, Grok.
The article does not say that this has happened. It says Democrats are asking agencies for answers because DOGE’s role, the use of proprietary tools, and the handling of sensitive data remain unclear. In that context, the question is whether the government can demonstrate that public data is not being used in a way that creates private advantage.
Rules Already Apply to Federal AI
Connolly writes that federal agencies are already “bound by multiple statutory requirements in their use of AI software.” The requests point chiefly to the Federal Risk and Authorization Management Program, which standardizes the government’s approach to cloud services and helps ensure AI-based tools are assessed for security risks.
They also cite the Advancing American AI Act. According to the source article, that law requires federal agencies to “prepare and maintain an inventory of the artificial intelligence use cases of the agency” and to “make agency inventories available to the public.”
For Democrats on the committee, those requirements matter because AI use inside government is not simply an internal technology choice. If agencies are using or considering AI systems, the requests argue that those uses must be documented, reviewed, and made transparent where required.
The issue becomes more urgent when AI is connected to workforce cuts. Automation may change how agencies analyze records or make recommendations, while fewer employees may remain to review outputs, challenge mistakes, or understand the limits of the software.
Known AI Tools and Agency Data
Documents obtained by WIRED show that DOGE operatives deployed a proprietary chatbot called GSAi to approximately 1,500 federal workers. The GSA oversees federal government properties and provides information technology services to many agencies.
A memo obtained by WIRED reporters warned employees not to feed the software controlled unclassified information. The article also says the departments of Treasury and Health and Human Services have considered using a chatbot, though not necessarily GSAi, according to documents viewed by WIRED.
WIRED has also reported that the United States Army is using software called CamoGPT to scan its records systems for references to diversity, equity, inclusion, and accessibility. An Army spokesperson confirmed the tool exists but did not provide more information about how the Army plans to use it.
The Department of Education is another focus. Connolly writes that the department has personally identifiable information on more than 43 million people connected to federal student aid programs. He says he is concerned that borrowers’ sensitive information may be handled by secretive DOGE team members for unclear purposes and without safeguards against disclosure or improper use.
The Washington Post previously reported that DOGE had begun feeding sensitive federal data from Department of Education record systems to analyze spending. The source article also says Education secretary Linda McMahon said Tuesday that she was moving ahead with plans to fire more than a thousand workers at the department, after hundreds of others accepted DOGE “buyouts” last month. The Education Department has lost nearly half of its workforce, which McMahon says is the first step in fully abolishing the agency.
Why Oversight Is the Core Issue
Connolly’s argument is that AI can create risks even when data is not improperly disclosed. He warns that the inputs and parameters used for analysis may be flawed, that errors may come from the software design, and that staff may misunderstand AI recommendations.
Those risks are especially important in government settings because AI outputs can influence decisions about programs, spending, records, and people. If the purpose of a system is unclear, it becomes harder to judge whether the tool is appropriate. If safeguards are unclear, it becomes harder to know whether sensitive data is protected. If oversight is unclear, it becomes harder to assign responsibility when something goes wrong.
That is why the requests focus on documentation and accountability rather than only on the existence of AI tools. Democrats are asking agencies to explain what is being used, why it is being used, whether it follows federal requirements, and how private data is protected.
Connolly’s warning is blunt: “Without clear purpose behind the use of AI, guardrails to ensure appropriate handling of data, and adequate oversight and transparency, the application of AI is dangerous and potentially violates federal law.”