A federal agency has 20 working days under Section 552 of Title 5 to decide a records request — when the period lapses the statute attaches no penalty to the agency and leaves enforcement to a lawsuit the requester must bring

What is happening?

The Freedom of Information Act, Section 552 of Title 5 of the United States Code, gives a federal agency 20 working days to make a determination on a records request. The statute allows an extension of ten more working days in unusual circumstances, on written notice to the requester. What the period governs is the decision on the request, not the delivery of the records themselves.

463,541 requests remained pending government-wide at the close of FY2025. The comparable year-end figures were 200,843 for FY2023 and 267,056 for FY2024, so the count rose in each of the two intervals those three figures cover. Requests received also rose, reaching 1,707,197 in FY2025, of which 1,635,055 were processed. Over the same span the processing workforce moved the other way, from 5,638 full-time equivalent staff in FY2024 to 4,823 in FY2025.

The same direction shows up department by department in 2026 reporting. As of 2026-03, the Department of Defense carried more than 30,000 pending requests for FY2025, the Department of Education about 4,570 at the end of FY2025, the Department of Housing and Urban Development about 1,092, and the Department of State 27,619 on a counting period the source does not state, with complex requests there averaging about 270 working days. As of 2026-08, the Office of Personnel Management reported a year-end FY2025 pending count above 1,600 and an average processing time for simple requests above 84 days.

When the deadline lapses, no penalty attaches to the agency by operation of the statute. The lapse lets the requester treat the request as constructively denied and go to an administrative appeal or to court, which places the cost of enforcement on the person waiting. The one personnel sanction the statute names runs through the courts first — Section 552 directs the Office of Special Counsel to investigate whether disciplinary action is warranted only after a court makes a written finding that the withholding was arbitrary or capricious. The Government Accountability Office reported in 2018, in GAO-18-235R, that no such referral was made across FY2008 through FY2016, and that requesters asked for one in six cases between 2013 and 2016 without a court making the finding.

The ombudsman office created by Congress states its own limits plainly. The Office of Government Information Services, housed at the National Archives and Records Administration, says it cannot compel an agency to release records, overturn an agency decision, or force an immediate response. It received 6,061 requests for assistance in FY2025 and closed 6,088.

Whose problem is this?

The body that decides whether a request is answered on time is the same body whose failure to answer produces the delay, and nothing outside that body starts a clock that runs out. The requester carries the enforcement burden.

Where does this problem end?

The boundary matters because the remedy here was not left undesigned. A deadline was written, a referral path was written, and an ombudsman was created, so the question is not what to build but why the built parts do not bind.

What is the state now, and what should it be?

Now

What it should be — the target is written into the statute rather than left to judgment. A determination is due in 20 working days, or in 30 where the agency gives written notice of unusual circumstances, and every request that sits longer than that is outside the standard Congress set. That is what makes this a question of enforcement rather than of missing rules. No source opened here names a target level for the pending count itself, or a date by which the government-wide figure is meant to fall, and none names a target for processing capacity.

How big is it?

463,541 requests were pending government-wide at the close of FY2025, against 1,707,197 received during that year. Those two numbers describe requests, not people, and the distinction is what keeps this document from stating an affected population.

One request is not one person, in either direction. A single requester may file many requests in a year, which would make the number of people smaller than the number of requests. A single request may be filed on behalf of many people, which pushes the other way. No source opened here supplies a factor that converts one into the other, so the affected-population module for this document records the count as not derivable rather than presenting a range built on an assumption no source supports.

The one composition figure available is narrow. At the immigration-benefits component of the Department of Homeland Security, 99 percent of requests in FY2018 were for the immigration file of the person asking. That is a single component in a single year, and no source opened here restates it government-wide, so it is not generalized here.

What the pending count does establish is direction. The year-end figure rose across FY2023, FY2024 and FY2025 while requests received also rose and processing staff fell, and each of those three movements is reported in the sources listed below rather than derived here.

Under what conditions does it arise?

1. The clock runs but nothing runs out. A lapsed determination period converts into a constructive denial, which unlocks an administrative appeal or a lawsuit. Both of those require the requester to act, and both take longer than the period that was missed. The statute gives the agency a deadline and gives the requester the job of enforcing it.

2. The one personnel sanction has a precondition the courts control. The Office of Special Counsel investigates only after a court makes a written finding that a withholding was arbitrary or capricious. GAO reported no referral across FY2008 through FY2016, including six cases between 2013 and 2016 in which a requester asked for one and no court made the finding. Whether any referral occurred after FY2016 is not established by any source opened here.

3. Demand and capacity moved in opposite directions in FY2025. Requests received rose to 1,707,197 while processing staff fell from 5,638 to 4,823 full-time equivalents. Agency officials quoted in 2026 reporting attribute their own backlog growth to staff departures, and that attribution appears in news reporting rather than in a finding by an oversight body. No independent causal analysis was found in this round.

4. Most of the volume sits in one place, and it is first-person volume. One department receives roughly 60 percent of all federal requests, and at its immigration-benefits component the requests were overwhelmingly for the file of the person asking as of FY2018. A government-wide average therefore describes that department more than it describes the rest of the government.

5. The instruments that exist describe, mediate and recommend. The ombudsman office states that it cannot compel release. The advisory committee issued eleven recommendations in 2026-07, seven of them directed at Congress, and they carry no binding force. GAO made four recommendations in 2024 on backlog-reduction planning and data reliability and reported them implemented as of 2025-08, while the year-end pending count for FY2025 came in higher than for FY2024. No source opened here links those two facts to each other.

What has been tried?

Two directions have been tried repeatedly — move the deadline, and add an office that helps without power to order. Neither direction attaches a cost to a missed determination, and the referral path that does exist has a precondition that no court in the examined span met.

What was found?

Why is it still unsolved?

Enforcement absent — the standard is written, the remedy is written, and the trigger that would make either bite is held by someone other than the party that was wronged.

The first part is that the consequence of lateness is a lawsuit the requester must bring. A statutory deadline whose only remedy is private litigation converts a legal duty into a matter of how much time and money the waiting party has. For a person requesting a file about themselves, that cost is almost always higher than the value of suing, so the deadline functions as guidance for the agency and as an obstacle for everyone else.

The second part is that the one sanction aimed at people rather than at agencies cannot start on its own. A court must first find that a withholding was arbitrary or capricious, and only then does the Office of Special Counsel take up the question of discipline. Across the span GAO examined, that finding was not made, including in the six cases where a requester asked for it. A sanction with a precondition that is never met is indistinguishable in practice from a sanction that does not exist, and the statute reads the same either way.

The third part is that the recorded history of this deadline runs toward relaxation rather than enforcement. When agencies could not meet the original ten working days, the response in 1996 was to make the period twenty. When the Act was reopened in 2016, the change narrowed the grounds for withholding and strengthened the ombudsman office without attaching a cost to lateness. Each of those steps is defensible on its own terms, and none of them put a consequence behind the clock.

The fourth part is that the failure produces numbers rather than events. A pending count is published once a year, it goes up, and nothing in the process treats that as a trigger. There is no threshold that opens an inquiry, no date on which an unanswered request becomes something other than an unanswered request, and no office that is required to act when the figure moves.

What observation would mean it is solved?

Candidates — (a) the government-wide year-end pending count falls for several consecutive fiscal years while requests received do not fall (b) the statute is amended so that a lapsed determination period carries a consequence that does not depend on the requester suing, and the record shows that consequence being applied (c) the agencies receiving the most requests report determination times inside the statutory period for the ordinary case, on a basis they publish.

(a) alone is weak. A pending count can fall because fewer requests arrive, because requests are closed on procedural grounds rather than answered, or because the counting basis changed. No opened source states whether the year-end figure counts every open request or only those already past the deadline, so a movement in that number cannot yet be read as a movement in how many people are waiting past the deadline.

(b) alone counts an instrument rather than an outcome. Referrals could begin without records being released any faster, and a sanction that is used a handful of times a year against a caseload above a million requests would change the statute without changing the wait. The observation has to be paired with a wait that shortens.

(c) alone hides the tail. One department receives roughly 60 percent of all federal requests, so a government-wide average mostly describes that department. Averages also conceal the complex track, where one department reported a mean near 270 working days as of 2026. The three candidates have to be read together, and each of them has to be read against the counting basis that produced it.

What is it connected to?

Fills with researchimmigration casework and benefits adjudication, where records requests feed other pending proceedings · federal workforce capacity and hiring policy · administrative law remedies where a statutory deadline lacks a private enforcement mechanism · state public records laws and their own deadline regimes · records management and electronic recordkeeping standards. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• What the year-end figure counts. No source opened here states whether 463,541 counts every request still open at the close of FY2025 or only those already past the determination period. The two readings give very different answers to how many people are waiting past the deadline, and this document holds to the narrower claim that the requests were pending.
• Why the count rose as much as it did in FY2025. No source opened here carries a causal analysis by an oversight body. Agency officials quoted in 2026 news reporting attribute their own increases to staff departures, and that is an attribution carried by a news outlet rather than a finding by an auditor.
• Whether any referral to the Office of Special Counsel occurred after FY2016. No opened source updates the GAO figure. The absence of an update is not a confirmation that the count remained at zero.
• How often the unusual-circumstances extension is used. No opened source gives the number of requests that received written notice of an extension, so the share of the pending count that is inside an extended period is unknown.
• Who is waiting, government-wide. No opened source gives the composition of federal requesters by category. The 99 percent first-person share is one component of one department in FY2018, and no opened source generalizes it.
• What the wait costs the person waiting. No opened source aggregates the downstream effect of a delayed response on other proceedings. One opened source describes a single requester experience at one agency.
• The agency distribution of mediation requests. No primary ombudsman page opened here gives the breakdown of FY2025 requests for assistance by agency; that breakdown appeared only in secondary summaries during this round.
• An on-record explanation from the office that compiles the reports. No opened source carries a statement from the Department of Justice Office of Information Policy addressing the FY2025 increase itself, as distinct from publishing the figures.

See the evidence

The primary government-wide report was not read. The Department of Justice summary that carries the FY2025 figures returned HTTP 403 to automated retrieval in this round, so 463,541 pending, 1,707,197 received, 1,635,055 processed and 4,823 full-time equivalent staff are taken from two opened secondary sources that cite that summary and agree with each other. The statute itself was opened directly, and it is the basis for every claim here about what the law requires and what it does not provide. Two GAO product pages were opened for the referral finding and the 2024 recommendations, and the National Archives pages for the ombudsman limits and caseload are statements by that office about itself. Department-level figures for 2026 come from two news outlets citing agency numbers, and this document reports them as such rather than as audited findings. Where the sources are silent the silence is left visible rather than filled — the counting basis of the year-end figure, the cause of the FY2025 increase, the referral record after FY2016, the use of the unusual-circumstances extension, and the government-wide composition of requesters are all unresolved here. Five rows above carry no URL because the research round could not open or identify the page, and one of them, the 1974 enactment detail, is corroborated in part by the 1996 signing statement, which was opened and which describes the ten-day period it replaced. This is a Path A output, a research-based definition, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range Not derivable

The reason and what is missing are listed under “What is missing” below