Independent desk

The Deep State Times

Institutional power. Public record. Official contradiction.

Official contradiction

FBI NSL nondisclosure figures sit uneasily beside DOJ transparency tallies

Public Inspector General findings, Freedom Act statistical tables, and FBI National Security Letter practice have long described gag orders as routine. Annual Department of Justice transparency reports present a narrower, later, and more aggregated picture. The record does not resolve how many orders remain in force.

The Times desk · October 2, 2026

FBI NSL nondisclosure figures sit uneasily beside DOJ transparency tallies

National Security Letters remain among the most frequently used compulsory tools in federal counterterrorism and counterintelligence work. Statute authorizes the Federal Bureau of Investigation to demand certain subscriber, financial, and communications records from companies without a court order in many circumstances. Accompanying nondisclosure requirements, often called gag orders, have been standard according to successive Department of Justice Inspector General reviews and congressional testimony in the public record.

Those Inspector General reports, issued after the post-2001 expansion of NSL authorities, described widespread use of nondisclosure language, incomplete or inaccurate internal tracking, and cases in which letters were issued outside proper legal predicates. The reviews did not purport to declassify operational files. They did document that gag provisions were attached as a matter of course in large volumes of letters, and that recipients were told they could not disclose even the existence of the demand.

Congress later required more public statistical reporting. The USA FREEDOM Act of 2015 directed annual transparency publications covering FISA orders, National Security Letters, and related nondisclosure. The Department of Justice and the Office of the Director of National Intelligence have since released tables that list, among other figures, the number of NSLs issued in a reporting year and, in some periods, how many of those letters included nondisclosure obligations.

The published tables are not a running census of gags still in effect. They generally count letters issued in a calendar or fiscal window. They do not, according to the face of the reports, tally how many prior-year nondisclosure requirements remain unlifted, how many have been modified after a recipient challenge, or how many companies remain bound by language first served years earlier. Recipients that later obtained partial unsealing through litigation appear in dockets, not as a complete overlay on the annual statistical annex.

FBI officials have said in public settings that nondisclosure is used when disclosure would risk an investigation, a source, or a method, and that the Bureau reviews gags and will lift or narrow them when the statutory standard is no longer met. Department of Justice filings in challenges under 18 U.S.C. § 3511 have similarly described a process of recertification and, in some cases, agreement to more limited secrecy. Those positions are paraphrases of litigation and oversight statements already on the public docket. They do not quantify the stock of live orders.

The numerical gap is structural. An annual issuance count can fall while the inventory of still-secret letters grows, if old gags are not systematically terminated. Conversely, a year with fewer new letters can coincide with a large inherited stack. Transparency reports that present “NSLs issued” and a binary or partial gag statistic for that year do not, on their own, answer how many entities remain forbidden from stating that they received a letter. Inspector General work from the mid-2000s through later follow-ups emphasized weaknesses in that kind of longitudinal tracking.

Court records add fragments rather than a total. Technology firms and libraries have litigated NSL gags; some opinions describe multi-year secrecy, delayed notice to customers, and government arguments that even aggregated disclosure would cause harm. Those cases are not a statistical sample of all letters. They show that gag duration can exceed the year of issuance, which is the unit most visibility reports use.

Budget justifications and intelligence authorization materials name the NSL program as an ongoing collection tool but do not publish a gag inventory. The Foreign Intelligence Surveillance Court oversees other authorities; NSLs of the classic ECPA and Right to Financial Privacy type generally do not require that court’s prior approval. That jurisdictional split helps explain why FISA statistical reports and NSL tables can be read side by side without producing a single reconciled gag number.

Civil liberties organizations have used the Freedom of Information Act to obtain redacted NSL forms, policy memos, and older statistical extracts. Productions often leave in place the same nondisclosure theory the letters themselves impose. The public file therefore contains both the government’s annual counts and a long paper trail in which gag language is treated as default. Those two genres of document are not designed to foot to the same total.

Nothing in the open budget, Inspector General, or docket record authorizes a claim that a classified master list has been obtained. What the record does show is a persistent mismatch of units: letters issued versus orders still binding; yearly tables versus multi-year secrecy; recertification described in principle versus an unpublished remainder. Until reports state the stock of active NSL nondisclosure requirements, and the number lifted that year, the published tallies will continue to under-describe the gag architecture the Inspector General already found to be widespread.

The Department of Justice’s annual transparency publications remain the principal official source for current-year NSL volume. They should be read with the older Inspector General findings and with the statutory design of NSL secrecy, which attaches at issuance and can outlast the reporting cycle. That is the comparison the public papers support. It does not require an unseen file. It requires noticing that issuance and duration are different facts, and that only one of them is regularly reduced to a table.