The WHO pandemic agreement was adopted in May 2025 with no legal force, and after seven annex negotiating rounds since September 2025 it still cannot open for signature

What is happening?

A special session of the World Health Assembly in December 2021 created an intergovernmental negotiating body to write a pandemic agreement. That body met 13 times in formal and resumed sessions between 2022 and April 2025 and completed a text. Its first deadline, the assembly of May 2024, was not met: as of 2024-05-24 convergence had been reached on 13 of the 34 articles, and the assembly moved the date forward by one year.

On 2025-05-20 the assembly adopted the text — 124 in favour, 0 against, 11 abstentions. One member state did not take part in the vote, its withdrawal from the organization being under way.

Adoption did not give the agreement legal force, and the design says so. The agreement can open for signature only after a further instrument, the annex on pathogen access and benefit sharing, has been concluded, and it enters into force only after 60 states ratify. The annex was moved off the main text and onto a separate track, to be negotiated by an intergovernmental working group.

That working group has met seven times between September 2025 and July 2026. Its sixth meeting, resumed in May 2026, ended without agreement, and member states agreed to extend the negotiating mandate. Its seventh meeting, running 2026-07-06 to 2026-07-17, closed with differences remaining, and an eighth meeting was set for 2026-09-14 to 2026-09-18. A draft annex circulated in March 2026 was not accepted by the African regional group, which asked for a return to the February text; the reported points of disagreement are mandatory registration, the binding force of the benefit-sharing obligations, and technology transfer, with member states of the European Union and the pharmaceutical industry reported on the other side of those three.

As of 2026-08-19 the deadline for concluding the annex stands at the assembly of May 2027, or a special session during 2026, with no opened source saying which applies. The agreement is not open for signature, so the count of ratifications is zero and the ratification process has not begun.

Whose problem is this?

The body that has to conclude the annex is the same body that has moved its own deadline more than once, and nothing outside it can compel the timetable. The people whose access the annex is meant to govern are not present in the negotiation as a constituency that can be counted.

Where does this problem end?

The boundary here is drawn around the absence of a concluded annex and not around the question of what that annex should contain.

That distinction is what keeps this document measurable. A reader who wants binding technology transfer and a reader who wants none of it can both agree on the dates below.

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

Now

What it should be

The target state here comes from the institution itself, which is unusual for this corpus, and the institution has stated it twice over. The agreement defines its own route to legal force: the annex is concluded, the agreement opens for signature, and 60 states ratify. The date for the first of those three is what has moved. The original date for the text was the assembly of May 2024 and it became May 2025. The date now attached to the annex is the assembly of May 2027, with a special session during 2026 named alongside it as an alternative.

The body that has to conclude the annex is the same body that set the date for concluding it, and nothing outside that body can compel the timetable.

How big is it?

The magnitudes the opened sources actually carry are institutional and probabilistic. None of them is a headcount of people harmed by the delay.

Institutional scale. The organization has 194 member states. 124 voted to adopt the text on 2025-05-20, with 11 abstentions and none against. Entry into force needs 60 ratifications, and none has been deposited because signature has not opened. Seven annex rounds have been held and an eighth is scheduled.

Exposure ceiling. World population was about 8.2 billion in 2024. That is the outer bound of who could be exposed to a future pandemic, and it is a ceiling and not a count of persons affected by the absence of the annex.

The historical analogue. As of July 2021, vaccination coverage in some low- and middle-income countries stood below 1 percent of population. That figure measures what happened under the arrangements in force at the time, and no opened source states how much of that gap this annex would have removed.

Reported risk, at low grade. A secondary report of a 2021 journal study gives an annual probability of an extreme novel epidemic of about 2 percent and a lifetime probability of about 38 percent. The journal page returned an automated refusal in this round, so the figures are carried only through the secondary report and are graded accordingly below. They are not multiplied by anything here.

No affected population is derivable, and the reason is structural. The mechanism that would generate a real count is the allocation of pathogen-derived products the annex is meant to establish, and that mechanism does not exist, so there is nothing to count against. Applying an annual probability to world population would produce an expected value that no source carries.

No opened source counts the people who lose measurable access while the annex remains unconcluded.

Under what conditions does it arise?

1. Legal force was made conditional on a document that had not been written. The text adopted in May 2025 cannot open for signature until the annex is concluded, and the annex was placed on a separate negotiating track rather than inside the text. So a completed vote and an unfinished negotiation coexist by design, and the adopted instrument sits at zero ratifications with nothing irregular having happened.

2. The hardest subjects were the ones moved to the separate track. Registration, the binding force of benefit-sharing obligations and technology transfer are the terms that decide who bears cost. Adoption of the main text in May 2025 was therefore reachable at a lower price than entry into force, and the difference in price is the annex.

3. Non-conclusion needs no coalition. Concluding an annex requires agreement among member states, and the absence of an agreed text is the one outcome that requires nobody to organize it. Two positions that are far apart on the underlying value question can decline the same draft for opposite reasons, and the result of that arithmetic is another meeting.

4. The deadline can be reset by the body that set it. The May 2024 date became May 2025, and the annex date now reads as May 2027 or a special session during 2026. No opened source identifies a default text, a fallback authority or a penalty attached to missing any of those dates.

5. The instrument actually in force covers a narrower class of pathogen. The 2011 influenza preparedness framework applies to influenza viruses, so a non-influenza pathogen of the kind that circulated from 2019 falls outside it.

6. Even a concluded annex binds only the states that ratify. Adoption is a vote and ratification is a separate national procedure, so the two counts do not have to match. No opened source estimates how many of the 124 adopting states would ratify.

What has been tried?

The text of the agreement is finished and the annex that would give it legal force is not.

Both deadlines that have fallen due so far were moved rather than met. The first belonged to the text and moved once. The second belongs to the annex and has now moved twice, from the sixth meeting to the seventh and from the seventh to a date that is written two ways.

What was found?

Why is it still unsolved?

Coordination failure — legal force was made conditional on a second document that had not been written, and nothing in the design attaches a consequence to not writing it.

The first part is that adoption and entry into force were separated, and the separation ran along the line where the money is. Registration, benefit sharing and technology transfer decide who supplies, who pays and who receives. Those three were carried onto a track of their own while the rest of the text went to a vote. So a majority large enough to adopt was assembled in May 2025 without the terms that determine what adoption is worth, and the instrument that resulted has 124 votes behind it and no ratifications in front of it.

The second part is that the absence of an agreed annex is the cheapest outcome to sustain. A conclusion needs agreement; a non-conclusion needs nobody. The states asking for binding obligations and the states asking for narrower ones can decline the same draft in the same week for opposite reasons, and neither has departed from the procedure by doing so. Seven rounds have ended that way, and the eighth is scheduled.

The third part is that missing the date changes the calendar and not the legal position of anyone. The assembly that set the deadline is the body that can move it, and it has moved it. No opened source identifies a default text that takes effect if the annex is late, a body that gains the drafting power instead, or any cost that lands on anyone when the date passes. A treaty that has been adopted and a treaty that is in force look the same in a headline and are not the same instrument.

The fourth part is that the delay produces no event to notice. Nobody announces that a negotiation has stopped, and none of the rounds has stopped in any formal sense. Each one ends with a statement that differences remain and a date for the next one, the earlier deadline is replaced by a later deadline, and the count of ratifications stays at zero without ever being reported as a failure.

What observation would mean it is solved?

Candidates — (a) the working group concludes the annex and the assembly adopts it (b) the agreement opens for signature and 60 ratifications are deposited, so that it enters into force (c) pathogens shared under the mechanism produce an allocation of the resulting products that actually reaches the states the agreement names.

(a) alone is weaker than it looks. It is the observation this document is built to notice, and it is also the one that can be satisfied at the lowest price. An annex can be concluded with registration that is voluntary, with obligations whose binding force is left to later instruments, or with timelines long enough that nothing moves inside a real outbreak. Conclusion is also not adoption, and adoption is not ratification.

(b) alone counts instruments rather than access. Sixty ratifications out of 194 members means the agreement is in force among those sixty. Nothing guarantees that the states which manufacture and the states which need are on the same side of that line, and no opened source estimates who ratifies.

(c) is the strongest and the slowest. An allocation reaching a named recipient is the only one of the three that shows the mechanism doing the thing it exists to do, and it cannot be observed until a pathogen event occurs. That places the confirming observation downstream of the event the whole arrangement is meant to prepare for.

Any one of the three can be true while the other two are false, so they have to be read together.

What is it connected to?

Fills with researchthe scope limit of the 2011 influenza framework as a template for what a broader instrument has to solve, regional manufacturing capacity for pandemic countermeasures, the design of conditional entry-into-force clauses in other multilateral instruments, and the treatment of intellectual property in medical countermeasure supply. Relation type and evidence grade were not established in this round.

What these sources do not say

• When the annex negotiation ends. Two dates travel together in the organization statements, the assembly of May 2027 and a special session during 2026, and no opened source says which one applies or what decides between them.
• How many states will sign or ratify. The adoption vote of 124 is not a count of ratification intent, because signature and ratification run through each national procedure separately, and no opened source offers an estimate of the number.
• What the eighth meeting will take up. The event listing carries the dates and no agenda document was posted at the time of this round.
• The positions in the words of the parties holding them. The account of the March 2026 draft rests on one reporting source. No opened source carries the statement of the African regional group in its own words, none carries the position of member states of the European Union in its own words, and none carries the position of the pharmaceutical industry in its own words. A search in this round did not locate the statement of the industry association that the reporting source summarizes.
• Whether the African regional group holds one position. Whether the reported position is the agreed position of all of its member states, or whether states inside it differ, is absent from every source opened here.
• Why the rounds have not concluded. Sources record that meetings ended with differences remaining and that mandates were extended. None carries an on-the-record explanation, attributed to any party, of what would close the remaining differences.
• The size of the access gap the annex would close. The coverage figure below 1 percent describes July 2021 under the earlier arrangements. No opened source states how much of that gap this annex would have removed, so the counterfactual that would size the problem is absent.
• The primary study behind the risk figures. The journal page returned an automated refusal, so the annual figure of about 2 percent and the lifetime figure of about 38 percent are carried only through a secondary report, and the counting basis of the study was not read.

See the evidence

No primary negotiating document was read. The pages of the organization itself were opened directly and carry the condition for entry into force, the extension of the annex mandate in May 2026, the outcome of the seventh meeting, the dates of the meetings, the scope of the 2011 framework and the July 2021 coverage figure. The UN population release was opened directly for the 2024 population figure. Everything else comes from news and specialist reporting that cites documents this round did not open, and one encyclopedia entry used only as a cross-check on the vote counts. Where sources overlap they agree: the adoption date and the vote counts appear identically in two opened sources, the May 2026 extension appears in two, and the fact that the agreement has no legal force until the annex is concluded appears in three. Where the record is unsettled it is left visible rather than resolved: the deadline for the annex is carried as the assembly of May 2027 in some places and as a special session during 2026 in others, with no opened source choosing between them, and the account of the March 2026 draft and of the disputed subjects rests on a single reporting source rather than on statements from the parties themselves. The journal study behind the probability figures could not be retrieved and its row is listed with the refusal recorded. 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