The Foreign Intelligence Surveillance Court has, since the USA FREEDOM Act, received written argument from appointed amici curiae when it considers novel or significant questions of law. Those briefs, unsealed in redacted form on the court’s public docket, repeatedly return to Section 702 of the FISA Amendments Act: targeting of non-U.S. persons reasonably believed to be abroad, incidental collection of U.S. persons, and subsequent queries of that repository by FBI, NSA, and CIA personnel.

The Office of the Director of National Intelligence, by contrast, publishes yearly Statistical Transparency Reports that tabulate 702 “targets,” “estimated” U.S. person queries, FBI “query terms,” unmasking requests, and related figures. Officials have described those tables as the public accounting of how often the authority is used. The two records sit in public view. They do not describe the same object in the same units.

Amicus submissions unsealed after the 2018, 2019, and later 702 reauthorizations, including filings associated with the court’s annual 702 certifications, have argued that “target” counts in the ODNI reports do not map onto the number of unique selectors or the volume of communications acquired. According to those briefs, a single target can generate many selectors; a selector can generate large volumes of traffic; incidental U.S. person communications are not separately enumerated as a headcount of people. The reports, amici have written, therefore cannot answer whether minimization procedures are adequate in practice.

ODNI’s published methodology, as restated in successive reports, treats a target as a person, group, or entity against whom collection is directed. Query statistics are presented as counts of U.S. person query terms used against 702 holdings, with FBI figures broken out after statutory changes that required more logging. The reports state that some query numbers are estimates, that certain systems cannot produce unique-person counts, and that unmasking figures reflect requests rather than unique identities. Inspectors general at DOJ and ODNI have, in unclassified summaries, similarly cautioned that query logs and targeting records are not a census of U.S. persons in the 702 corpus.

The court’s public orders, including those summarizing compliance incidents, have alleged repeated FBI querying of 702 data using U.S. person identifiers without the required predication, later addressed by training, system prompts, and statutory query restrictions. Amici have treated those incidents as evidence that statistical query totals, even when rising or falling year to year, do not disclose the legal character of the searches. ODNI tables show year-over-year changes in FBI U.S. person query counts after the 2018 and 2024 reauthorization cycles; they do not, the public reports say, classify queries as compliant or noncompliant.

Procurement and budget documents add a third, still incomplete, layer. Intelligence Community budget justifications submitted to Congress, in unclassified form, describe 702 as a high-volume signals intelligence program whose costs sit inside NSA and FBI information-technology and analytic accounts. Those justifications do not publish a unit cost per target or per query. Amicus briefs have asked the court to require the government to explain how volume and cost relate to necessity. The government’s public responses, as paraphrased in FISC opinions, have held that necessity is a legal determination under the statute and the Fourth Amendment, not a statistic the transparency report is designed to supply.

Congressional reauthorization debates have cited both sources. House and Senate intelligence and judiciary reports have pointed to ODNI tables as proof of scale and to FISC compliance summaries as proof of risk. Civil-liberties groups that have served as or supported amici have argued in public comments that the statistical reports omit the incidental U.S. person collection that the court must still minimize. ODNI and DOJ public statements have maintained that naming U.S. persons in 702 reporting would itself raise privacy and operational concerns, and that the existing counts are what current systems can produce without further collection on Americans.

The docket therefore shows a stable mismatch. Amici write in the language of particularity, incidental collection, and query purpose. The transparency reports write in the language of targets, query terms, and unmasking requests. Inspectors general have documented gaps in logging that make a person-level 702 census unavailable. Neither the unsealed amicus briefs nor the ODNI reports claim to close that gap. The record, as published, is two ledgers for one authority, kept in different units, both incomplete by their own stated methods.