The Department of Housing and Urban Development is withholding documents about the Department of Government Efficiency's use of artificial intelligence to inform policy decisions, citing a privilege that federal law does not recognize. The agency's denial came in response to a Freedom of Information Act request by Democracy Forward, a nonprofit legal organization, and raises questions about transparency in AI-driven policymaking.
More than 100 documents that had been requested about HUD's AI use for decisionmaking were withheld. Among the reasons HUD gave for not releasing them were a nonexistent "AI privilege" and a presidential communications privilege that courts generally apply only to the president and immediate advisers. Several of the withheld document titles - including "GPT defined Econ Analysis approach 11 10 25.docx" and "RegulatoryAnalysisPrompt.pdf" - appear to indicate DOGE staff at HUD were using AI tools to help shape policy.
How the FOIA denials were framed
Nearly all the withheld documents carried an Exemption 5 label, which protects "deliberative process privilege" - the internal back-and-forth between federal workers giving candid feedback on draft policies. But in a handful of cases, HUD's FOIA office went further, listing "draft of AI prompt" and "deliberative AI input" as specific reasons to deny access.
John Davisson, deputy director of enforcement at the Electronic Privacy Information Center, said the deliberative process privilege is meant to encourage federal employees to speak honestly. "There is no AI exemption under FOIA," he said. "AI systems, computers are not entitled to candor." One document, "DFR Template_Workflow Prompt Directory (3).pdf," was also withheld under "presidential communications privilege," which Davisson said raises "questions about where the prompts may have come from."
Risks of invisible AI in rulemaking
Tori Noble, a staff attorney at the Electronic Frontier Foundation, said the lack of transparency is especially concerning because AI tools can hallucinate, show bias, or get facts wrong. "Having access to the prompts is really the best way to be able to tell what officials are using these tools for and how harmful those uses might be," she said.
Mark Fagan, a lecturer at the Harvard Kennedy School, noted that while some AI use could be considered part of an internal deliberative process - similar to a Google search - the current development stage of AI tools calls for disclosure. "If AI is being used to assess policy as one of the tools in the toolkit, it is good protocol to indicate that," he said, "in part to try and build confidence in the use of AI in government."
Dan McGrath, senior oversight counsel at Democracy Forward, said the withholding of AI input as privileged "raises serious questions." He added: "If the government is going to use AI in formulating policy that affects us all, the public has a right to understand its impact. Existing privileges are meant to ensure that public officials share their candid views, not to shield the impact of AI on our government."
Why this matters for government professionals
There are no U.S. laws requiring agencies to disclose when AI has been used to create or modify rules. As agencies explore AI for Government functions, the HUD case signals that the current FOIA framework may not be equipped to handle AI-driven decision-making. Policy professionals who want to understand the technology's role in regulatory work can follow an AI Learning Path for Policy Makers that covers the capabilities and risks of these tools. The withheld documents leave open the question of whether AI was used to flag regulations for rescission - and if so, how reliable that process was. For those inside government, the episode underscores the need to build clear internal protocols for AI use before transparency disputes become the norm.
Your membership also unlocks:
