NSA Records Languish at National Archives for Now
Last year, the National Archives (NARA) acquired a large number of historically valuable National Security Agency records. But they remain inaccessible to researchers, at least for the time being.
David Langbart of NARA described the situation at a closed meeting of the State Department Historical Advisory Committee late last year. According to recently published minutes of that meeting:
“The [NSA] records consist of approximately 19,000 folders without any real arrangement. These records mostly consist of technical, analytical, historical, operational, and translation reports and related materials. Most of the records date from the period from the 1940s to the 1960s, but there are also documents from the 1920s and 1930s and even earlier. The NSA reviewed the records for declassification before accessioning and most documents and folder titles remain classified. Langbart concluded that the finding aid prepared by NSA was the only practical way to locate documents of interest for researchers, but it is 557 pages long and is classified.”
The National Archives confirmed that this description remains accurate today.
So not only are these thousands of half-century-old records still classified or otherwise unavailable, but the finding aid that would enable researchers to locate specific documents of interest is itself a classified document.
The Federation of American Scientists asked NSA officials to voluntarily declassify the 557-page finding aid as a first step towards making the NARA collection useful to researchers.
They agreed to do so.
“We can have a redacted version for you by September,” wrote Dr. David J. Sherman of NSA. “We of course will provide one to NARA as well.”
Dr. Sherman noted that the collection includes documents of widely varying complexity. “Judging by their titles, some almost certainly require significant training in mathematics and engineering to understand. Others appear to have been written for more general audiences.”
Furthermore, although the collection as a whole is maintained as classified, “just under one third of the folders appear to be unclassified in full,” he estimated.
Under the circumstances, classifying the entire set of records along with its descriptive catalog was obviously not optimal, he agreed.
“I take the point about this foreclosing any possibility for researchers to know what might be available in the collection and agree it is something we should have addressed in this instance and need to fix in the future,” Dr. Sherman said.
Therefore, he added, “in any similar situations in the future — i.e. ones where we are transferring large, mixed collection such as this — we’ll make it standard practice to consider whether the percentage of unclassified materials is high enough to provide NARA with a redacted finding aid at the time of the transfer.”
These ideas aim to advance the detailed policy solutions needed to foster public trust and implement fairness in the adoption of AI across diverse domains, from healthcare and government benefits to rural access, education, and worker protections.
The evidence is clear: algorithmic pay-setting is established in app-based work, and payroll/timekeeping failures show how software can produce systemic wage harm at scale
While a few states have taken steps to implement decision-making mechanisms for certain AI systems, too many leaders are simply accepting narratives about AI’s purported public benefit at face value – jumping to the “how” of AI implementation before thoroughly vetting potential systems and deciding whether they are appropriate to use at all.
When properly structured — with specific numeric targets, secured financial obligations, independent monitoring, and meaningful enforcement — CBAs transform data center deals into durable community partnerships.