Summary

  • GAO examined 311 international agreements and qualifying nonbinding instruments reported to Congress from October 2023 through March 2025. Ninety were late at one or both reporting points; the late items averaged 2.3 reporting cycles behind.
  • State's Treaty Office lacked a systematic timeliness record for incoming agency submissions. Its intake for qualifying nonbinding instruments omitted the legal-authority information Congress was meant to receive, and its monthly tasker did not seek an all-or-none response from every source.
  • State rejected GAO's recommendation for periodic completeness confirmations and said written publication procedures already existed. GAO kept both recommendations open. The test is an auditable chain from each agency's submission or zero report to the two congressional notices and public posting.

A clock with two starting guns

The GAO audit of international commitments begins with a problem Congress tried to solve in 1972: the executive could conclude agreements before legislators learned what had been agreed. Amendments effective in September 2023 widened the rule beyond agreements binding under international law. Qualifying nonbinding instruments, which can carry serious foreign-policy effects without creating those legal obligations, also entered the disclosure system.

Each commitment now has two reporting moments. State must tell Congress after signature, conclusion or finalization, then again after the agreement enters into force or the instrument becomes operative. The first report lets Congress see a commitment while its consequences may still be developing; the second records the point at which it begins to work. An agency must send State the text and legal authority within 15 days of conclusion. State reports monthly. For items subject to public disclosure, text and legal authority must appear on State's site within 120 days of taking effect or becoming operative.

These clocks are useful only if the receiving office knows a file exists.

GAO measured 311 agreements and instruments that State did report to Congress from October 2023 through March 2025. Of those, 221 were on time and 90 late at one or both moments. The late group averaged 2.3 reporting cycles behind. International agreements had a 30% late rate; qualifying nonbinding instruments, 25%. One agreement concluded on 4 November 2024 was due in Congress's hands by the end of December. It arrived at the end of March 2025, about five months after conclusion. The difference between an early notice and a spring retrospective can be the difference between oversight of a live choice and an autopsy of it.

State could not time the handoff it received

The Treaty Office collects the commitments from other agencies and State's own bureaus. It could manually reconstruct which agencies were associated with delayed files, but it did not systematically track how promptly each source sent material to the office. That makes blame by anecdote easy and control by measurement hard. A delay might occur before the handoff, at the Treaty Office, or at both. Without the dates, management cannot locate the repeat offender or distinguish a difficult case from an indifferent one.

The form itself had a hole. For qualifying nonbinding instruments, State's monthly intake requested title, dates, participants and publication details, but did not ask agencies for the legal authority Congress was supposed to receive. State had a process for identifying authority for binding agreements. Its nonbinding route used broad Article II language in the congressional list while omitting source-specific authority collection. The result is a disclosure record that can name an instrument yet leave the reader without a precise account of the power used to make it operative.

The most revealing gap is an empty inbox. State's tasker asked offices to respond when they had a qualifying instrument. It did not ask each to confirm that all applicable commitments had been sent, or that none existed in that period. GAO recommended periodic all-or-none confirmations. State called that unreasonable and did not concur. The distinction is more than paperwork. A positive-only system counts the files it receives. A zero-report system also identifies which sources were asked and which answered. GAO expressly did not independently inventory every potentially reportable commitment, so the 311 cannot be treated as the universe.

An independent agency-by-agency inventory is the missing fact that would settle how complete the stream was.

Congress had also required a way for State staff to flag suspected or known failures to report. GAO found no designated mechanism and no clear instruction that the tasker could serve that purpose. The system had a route for ordinary submissions and no unmistakable route for saying that an ordinary submission was missing.

Public disclosure after the public moment

GAO's separate sampling analysis shows why a congressional list alone is insufficient. From a generalizable sample of 156 reported items, GAO estimated about 75% were subject to the public-posting rule. Among that eligible group, it estimated State published the text and legal-authority information about 86% of the time, but met the 120-day deadline only about 53% of the time. These are sample-based estimates, with exemption rules and uncertainty around them; they are not counts obtained by multiplying 156 by a headline percentage.

Even a posted item could require detective work. GAO found the texts and authority details spread across four webpages, with no single search across country, agency or subject. A document can be online and functionally elusive. State disputed GAO's call for written operating procedures, saying it already had them; GAO drew a line between regulations stating a 120-day duty and operational steps that reliably meet it. State accepted the website-search recommendation. On the GAO recommendation page, all seven items were still open when checked on 9 November 2026. Open is GAO's recorded status, not a claim about every action State may have taken since the audit.

The deeper choice is whether the executive's disclosure system is a tray or a ledger. A tray can show what reached it. A ledger records the sources expected, their positive or zero responses, the legal authority attached, every handoff date and the moment the public can find the item. Congress built two clocks because time changes what oversight can do. State must make the clocks belong to the whole chain.

Sources