The U.S. Department of Housing and Urban Development (HUD) has blocked over a hundred AI‑related policy documents from a Freedom of Information Act request, invoking an invented “AI privilege.” The move raises serious concerns about transparency and accountability in government use of artificial intelligence.

मुख्य बिंदु (Key Takeaways)

  • HUD denied FOIA access to more than 100 documents concerning AI‑assisted policy work.
  • The denial was justified by a fictitious AI privilege and a presidential‑communications privilege.
  • Experts warn that lack of AI transparency can lead to bias, hallucinations, and reduced public trust.

The Department of Housing and Urban Development’s internal unit, the Department of Government Efficiency (DOGE), has been experimenting with artificial‑intelligence tools to shape policy. When Democracy Forward, a nonprofit legal group, filed a Freedom of Information Act (FOIA) request for those records, HUD responded by invoking a non‑existent “AI privilege” and, in some cases, a presidential‑communications privilege that traditionally applies only to the President and close advisers.

Background and Key Players

Last year WIRED reported that Christopher Sweet, then a third‑year economics student at the University of Chicago, and Scott Langmack, a former PropTech startup employee, joined HUD’s DOGE team. Their mission was to use AI to flag agency regulations that might be rescinded or contracts that could be cancelled—part of a broader government‑wide effort to automate rule‑review. Sweet graduated in June, while Langmack now serves as the Executive Director of Deregulation AI at the Office of Management and Budget (OMB).

FOIA Request and Withheld Documents

Democracy Forward sought more than 100 records on HUD’s AI‑driven decision‑making. Among the withheld files were “GPT defined Econ Analysis approach 11 10 25.docx” and “RegulatoryAnalysisPrompt.pdf,” both marked as “deliberative AI input” or “AI prompt.” HUD cited Exemption 5—deliberative process privilege—to keep the documents sealed, suggesting they were part of internal policy‑development discussions before final decisions.

Expert Concerns and Legal Gaps

Electronic Frontier Foundation staff attorney Tori Noble warned that AI tools can hallucinate, embed bias, or simply produce incorrect outputs, making transparency essential. Harvard Kennedy School lecturer Mark Fagan argued that when AI is used as a policy tool, agencies should openly disclose its role to build confidence. Yet, no U.S. law currently obliges federal bodies to reveal AI involvement in rule‑making.

Implications and Way Forward

The reliance on a fabricated “AI privilege” highlights a regulatory vacuum around governmental AI use. As AI becomes more embedded in policy formulation, Congress and agency leadership will need to craft clear FOIA‑compatible guidelines that balance the need for candid internal deliberation with the public’s right to know how algorithms influence decisions.