Leopold v. Central Intelligence Agency, No. 16-cv-1833 (JMC), D.D.C. (July 21, 2026), full opinion (PDF)
Jason Leopold and Ryan Noah Shapiro sued the CIA under the Freedom of Information Act for records a Senate committee had cited in its report on the agency's detention and interrogation program. The CIA moved for summary judgment on the ground that it had searched for those records adequately. To win that motion an agency has to describe under oath what it searched, who searched it, and how.
The CIA gave that account for the database it searched first. For the four offices it turned to afterward, the agency said only that knowledgeable personnel had run targeted searches. On July 21 Judge Jia M. Cobb accepted the account of the database. She rejected the account of the four offices and denied the CIA summary judgment on the adequacy of its search.
What Happened
The request attached an index listing over 800 citations to the footnotes of the committee's declassified report. Eleven of the cited records the CIA never located.
CIA staff had identified a database called the Rendition, Detention, and Interrogation Network, or RDINet, as the location reasonably likely to hold responsive material, since committee staffers had drawn on that same repository. Search experts queried it using terms appearing in each document's title or, for cables, the cable citations. When a record did not surface there, reviewers moved to four offices where it likely originated. The characterization of those four searches arrived in a supplemental declaration filed during the litigation.
The Court's Analysis
The court held the CIA to declarations that "specify what records were searched, by whom, and through what process". Judged against that test the RDINet account survived, because it named the database queried and the terms run against it.
Naming four offices told the court nothing about how anyone had looked inside them. Those descriptions, Judge Cobb found, "do not denote which files were searched, or by whom, do not reflect any systematic approach to document location". The court measured them against National Security Counselors v. CIA, where comparable affidavits had already been held insufficient. The CIA's descriptions were thinner than the affidavits in that case. They never identified what records systems the four offices kept or how those systems were searched.
In its briefing the agency tried to repair the gap. The requesters argued that some of the missing records existed in paper or archival form, or in the personal files of retired employees. The CIA answered that its office searches had covered the archives and the former employee's files. The court noted that "these facts are stated in the CIA's summary judgment briefs without citation to the record", and read the late addition as confirmation that the original description had been deficient.
The requesters lost their own attack on the search terms. A requester cannot dictate the terms an agency runs, and a court will not second-guess reasonable terms because better ones can be imagined. Searching by the words in a document's title was reasonably calculated to find the records sought. Broadening the terms would only have produced more documents.
Neither side won summary judgment on the search. The court denied both motions without prejudice and pointed the agency toward a more detailed declaration or a supplemental search. Judge Cobb separately resolved most of the parties' disputes over the withheld records.
Why It Matters
Give every location searched the treatment the main repository gets. Name the records systems inside each one and the method run against them. The CIA could not tell the court what systems those four offices even maintained, which left no way to judge whether searching them was reasonable.
Put that detail in the sworn record. Assertions about scope that live only in a brief carry no weight. Reaching for them late invites a court to read the original description as deficient.
Reasonable search terms survive a challenge built on better ones. The question a court asks is whether the terms were calculated to find the material, not whether a sharper list existed.
A thin description is curable. Both denials ran without prejudice, and the agency can still prevail by filing a fuller declaration or by running the search again and describing it properly.
The full opinion is available as a PDF.
