Artificial Intelligence government transparency

NARA’s 2024 AI and Open Government Panel: What It Covered

What NARA’s March 2024 panel said about AI, public access and government transparency, with the original event report and its limits.

NARA’s Sunshine Week panel on artificial intelligence took place on 14 March 2024. It discussed public access, searchable records and agency uses of AI. This is a dated event summary and a research guide, not an announcement of a new FOIA rule.

Correction, 7 October 2026: This article has been revised against NARA’s official event account. Unsupported claims about legislation, agency programmes and quotations have been removed.

What the 2024 panel covered

The official NARA report, published on 15 March 2024, identifies Pamela Wright as moderator and participants Gulam Shakir of NARA, Abigail Potter of the Library of Congress, Eric Stein of the State Department and Bobak Talebian of the Justice Department.

The discussion addressed digitisation, metadata, optical character recognition, privacy detection and tools for finding or routing requests. NARA described experiments to improve catalogue searches; the State Department discussed its reading-room search; DOJ discussed the FOIA Wizard. These are distinct applications, not one shared system that automatically decides every request.

How to use the sources

ResourceUseCheck
NARA event accountIdentify the panel and its context.Keep the 2024 date explicit.
National Archives CatalogueLocate descriptions and digitised holdings.Record the catalogue identifier and inspect original pages.
FOIA.govFind agencies and request guidance.Confirm the receiving component and its instructions.

Start with the National Archives Catalogue for archival material or FOIA.gov for agency guidance. An archival search and a FOIA request are different routes. Search-result text may help locate a record, but the original document should support a quotation or factual claim.

Questions that remain useful

For an agency AI tool, ask what task it performs, which records it processes, how errors are reviewed and whether a human makes the final disclosure decision. Keep a search failure distinct from proof that no records exist. Where an OCR result is unclear, compare it with the scanned page.

The panel provides historical context for those questions. It does not establish that AI guarantees faster delivery, overrides exemptions or removes the need for human review.

Leave a Reply

Your email address will not be published. Required fields are marked *