DOGE Used AI for Housing Policy. The Government Doesn’t Say What

Members named Department of Government Efficiency (DOGE) who were working Department of Housing and Urban Development (HUD) used artificial intelligence to inform political decisions. Now, the agency appears to be refusing Freedom of Information Act requests for information on the development and use of AI tools, and how they have informed policy decisions, according to documents requested by FOIA by Democracy Forward, a non-profit advocacy group.

Last year, WIRED reported that Christopher Sweet, then a third-year student at the University of Chicago, joined the DOGE team at HUD, along with Scott Langmackwho came to DOGE from a domestic technology called Kukun. Sweet’s main goal, according to HUD officials who spoke to WIRED at the time, was to use artificial intelligence to identify agency rules that could potentially lead to refunds, or cancellations of contracts, such as part of a similar effort throughout the state.

At the time, HUD staff told WIRED that staff were being contacted to comment on the AI-proposed rules to remove them. Some workers, however, described the efforts as unnecessary.

Sweet graduated from the University of Chicago in June with a degree in economics; Langmack is now the director of AI deregulation at the Office of Management and Budget (OMB), under the Executive Office of the President, according to his LinkedIn profile.

More than 100 documents requested by Democracy Forward regarding HUD’s use of AI in decision-making were redacted. Among the reasons HUD cited for not releasing the documents were the non-existent AI access and access to presidential communications that is real but only available to the president and his closest advisers. Several declassified documents, whose names are shared in the FOIA but whose contents are unknown, appear to indicate that DOGE’s team at HUD is using AI tools to help make decisions.

Sweet, Langmack, HUD, OMB, and the White House responded to requests for comment.

One document, titled “GPT defined Econ Analysis approach 11 10 25.docx,” which belonged to Langmack, was not included in the FOIA because it was labeled an “intentional AI input.” Another document, titled “RegulatoryAnalysisPrompt.pdf,” also belonging to Langmack, appears to indicate that the DOGE team is looking to create regulatory analysis prompts. Several other documents that were withheld from being part of the negotiations were listed as “legal analysis” for various HUD programs, though it is unclear whether AI was used in the creation.

Tori Noble, a staff attorney at the Electronic Frontier Foundation, says the lack of transparency about how AI tools can be used to create or change policy is troubling, because the tools are known to be strong, show favoritismor more clearly change things. “We don’t always know how the tools are being used,” he said.

There are currently no laws in the US that require the government to disclose whether AI has been used to create laws, policies, or regulations.

“If AI is being used for policy testing as one of the tools in the toolkit, I think that at this point in the development and use of AI, it’s a good way to show that,” says Mark Fagan, a professor at the Harvard Kennedy School. “In part to try and build confidence in the use of AI in government.”