Who Becomes the Tenth Secretary-General?

The real race · 30 July 2026

One job, and a first count

António Guterres’s second term ends on 31 December 2026. By mid-September, eight declared candidates wanted to succeed him — five women and three men, six from Latin America and the Caribbean, whose regional “turn” it is by informal rotation, and two from Africa.

The Security Council chooses by straw poll: each of the fifteen members marks every candidate “encourage”, “discourage” or “no opinion” on a secret ballot. On 30 July, Rebeca Grynspan of Costa Rica led with ten encouragements; Carolyn Rodrigues Birkett of Guyana followed with nine. Two candidates therefore cleared the nine votes a recommendation needs — but on undifferentiated paper, so nothing was learned about where the five permanent members stood.

The device is older than most of the people using it. Olara Otunnu, then Uganda’s ambassador and now himself a candidate, invented the straw poll as Council president in 1981 to break a sixteen-ballot deadlock.

See the reported ballots →

The real race · 21 August 2026

The field gets worse, not better

Three weeks later the Council voted again, this time on eight names. Nobody reached nine. Rodrigues Birkett slipped from nine encouragements to eight and still moved into first place, because Grynspan fell from ten to seven. Rafael Grossi, the head of the International Atomic Energy Agency, kept his seven supporters while his discouragements tripled from two to six — the clearest hardening of opposition in the field. Michelle Bachelet collapsed to two. Ivonne A-Baki, nominated by Tonga four days before the vote, drew no encouragement at all and eight discouragements.

Every one of those discouragements was still anonymous. Until the Council switched to colour-coded ballots — one colour for the permanent five, white for the elected ten — a “discourage” could come from anyone, and reading the counts as veto threats was a category error. That switch had no fixed date. It was the hinge of the race.

These two counts matter for what follows, because the briefing document our simulated delegations were given carries them. When a persona argues about who can win, it is arguing from these numbers.

See the reported ballots →

The real race · as of 17 September 2026

What has to happen before 1 January

The Council’s third straw poll was due on 18 September, under France’s presidency — the day after our online event with University College Dublin. This snapshot was taken before it, so none of the chapters reflects it. More undifferentiated rounds were expected through September and October; then, at a moment of the Council’s own choosing, the ballots would be coloured and a permanent member’s “discourage” would become a de-facto veto signal.

To be recommended, a candidate needs nine encouragements and no permanent-member discouragement on coloured paper. The Council then adopts a resolution naming one person, and the General Assembly appoints — by acclamation every time since 1971. If nothing is agreed by 31 December, the precedents are a temporary extension of the incumbent (1966) or an acting Secretary-General (1961). Both are possibilities, not plans.

• Regional rotation and gender are political forces, not rules: 2016 proved a strong regional claim can lose.
• There is no cut-off for new candidates. A-Baki entered four days before a poll; a deadlock could produce a late compromise name.
• Seven elected members acting together can block anyone — fifteen minus seven leaves eight, one short of nine.

The simulation

The rooms we simulated

Between 14 and 17 September we put 5 questions to 6 simulated Security Council rooms. The first four were presented at the 17 September event; the two budget rooms were run the same week. Each one can be opened in full (free sign-in to TheDiplomat.AI required).

The participants are AI personas of the foreign ministers of the five permanent members — Marco Rubio, Wang Yi, Sergey Lavrov, Ed Miliband and Jean-Noël Barrot — and of the ten elected members serving in 2026: Bahrain, Colombia, the Democratic Republic of the Congo, Denmark, Greece, Latvia, Liberia, Pakistan, Panama and Somalia. Each persona argues from its own capital’s public record. The rooms come in 3 sizes: the permanent five alone, the elected ten alone, and all fifteen seats together.

Every room ran on the same settings, so the differences between them come from the question and not from the dial settings: the same briefing document attached (a 98.5-kilobyte file on the race — the rules, the reported poll figures and a dossier on each candidate, current to 24 August 2026); a neutral stance, meaning no behavioural steer towards consensus or towards hard bargaining; the strict agreement rule, under which a rewritten proposal has to be re-approved by every delegation before it counts as agreed; and one run of the same model at a low reasoning effort.

• The veto mapThe permanent five · 14 September →
• First and second choice, the permanent fiveThe permanent five · 14 September →
• First and second choice, the elected tenThe ten elected members · 14 September →
• The three finalists in the full CouncilAll fifteen seats · 15 September →
• The toughest budget negotiatorThe permanent five · 14 September →
• The same budget question, finalists onlyThe permanent five · 17 September →

Each room opens on TheDiplomat.AI with every turn, the proposals and who agreed to what. Reading it needs a free account: sign in with Google and the room opens straight after.

A room runs like a consultation: one delegation opens with a proposal, the others answer in turn — agreeing, amending or refusing — until every open point is settled or the turn limit is reached. Five of the six rooms settled every point. The sixth, the fifteen-seat one, ran out of turns; that is said plainly where it matters.

See the six rooms, with the questions as typed →

Room 1 · the permanent five · 14 September

The veto map

Asked to name, in one line each, the candidate each permanent member would most likely veto, the room answered in five turns: China tabled the whole list in the first turn and every other delegation confirmed only its own line.

The United States named Michelle Bachelet, citing her human-rights record and the congressional opposition reported in the briefing. China and Russia both named Rafael Grossi — Beijing over his agency’s reporting on the AUKUS submarine safeguards, Moscow over its reporting on Iran. The United Kingdom named Macky Sall on accountability grounds: protest deaths, jailed journalists and the 2024 amnesty law. France named Ivonne A-Baki, reading her drift from the two-state formula and her campaign against the veto itself as disqualifying. Five delegations, 4 different names.

“I assess Sall as my government’s strongest potential veto concern because of the accountability issues alleged in the supplied dossier, not as an announced British veto decision.”

Two things about this list should temper how much weight it carries. First, it is stable for an uninteresting reason: the same five lines came out of the consensus-seeking and the hard-bargaining versions of the room on the same day, because every delegation said its entry came from the supplied dossier. The room is echoing the briefing rather than reasoning past it. Second, the same personas answered differently in one-to-one chat that day — China’s named Bachelet for its own capital, and Russia’s named A-Baki.

No delegation used the word veto as a decision. Each called its entry an assessment, and each gave the same reason: the real ballots so far are undifferentiated, so nobody — inside the simulation or outside it — can say which member cast which discouragement.

“I identify Grossi as my government’s assessed veto risk for this negotiation, not an announced veto decision. Independent recent evidence confirming candidate-specific veto intentions was unavailable to me.”

See the five lines and the reasons given →

Room 2 · the permanent five · 14 September

First and second choices: the permanent five

Asked for a table of first and second choices, the five filled it in row by row. Nobody filled anyone else’s row, and nobody proposed a bridge candidate.

The United States put Rafael Grossi first for nuclear verification and institutional rightsizing, Olara Otunnu second. China put María Fernanda Espinosa first for sovereign equality and development. Russia and France both put Macky Sall first — Moscow for sovereignty and balanced engagement with the permanent members, Paris for executive experience and African engagement. The United Kingdom put Rebeca Grynspan first for mediation experience and institutional independence. That is 4 different first choices among five delegations, and Carolyn Rodrigues Birkett as the second choice of 4 of the five — nobody’s favourite, almost everybody’s fallback.

“I propose a delegation-owned table of assessed first and second preferences, not a joint nomination or an attribution of secret ballots.”

Across the twelve permanent-five rooms and chats that asked this question on 14 and 15 September, two rows never moved: the United States on Grossi and Otunnu, Russia on Sall and Rodrigues Birkett. China and the United Kingdom keep the same pair of names every time and only swap the order — Beijing between Espinosa and Rodrigues Birkett, London between Grynspan and Rodrigues Birkett. France is this run’s exception: in every other run of the same question Paris put Grynspan first.

“My acknowledgment of other delegations’ rankings is not an endorsement of their candidates or a commitment concerning any ballot or veto.”

See the table and the reasons given →

Room 3 · the elected ten · 14 September

First and second choices: the elected ten

The same question, put to the ten elected members, produces a far more concentrated answer than it does among the permanent five.

Rebeca Grynspan is the first choice of 5 of the ten and the second choice of 5 more — she is in every delegation’s top two. Carolyn Rodrigues Birkett is first for 3 (Pakistan, Liberia and Somalia) and second for 4. What these ten reward, in their own words, is mediation under the Charter, the protection of civilians and humanitarian access, accountability, women’s participation, and reform that does not cut into delivery on the ground.

“I rank Carolyn Rodrigues Birkett first for the small-state perspective, Charter-based diplomacy and Palestinian civilian protection described in the supplied dossier.”

Two delegations look elsewhere. Bahrain puts Rafael Grossi first, for nuclear verification and crisis diplomacy. The Democratic Republic of the Congo puts Olara Otunnu first, because the briefing records his attention to the DRC, the Washington Accords and children in armed conflict. Both still place Grynspan second, which is why the concentration survives the exceptions.

“I record Olara Otunnu as my provisional first choice because the supplied dossier identifies explicit attention to the DRC, support for the Washington Accords and experience protecting children in armed conflict.”

This room is the most repeatable of the six. Across the six large-room runs of 14 and 15 September, seven of the ten named the same first choice every time; Pakistan and Panama did so in five of the six; Liberia is the one that moves, alternating between Grynspan and Rodrigues Birkett. Every one of the ten keeps one of those two women in its top two in every run.

See the ten rows and the tally →

Room 4 · all fifteen seats · 15 September

The finalists in the full Council

The fourth room builds on the veto map. Its question sets aside the four candidates the veto room had named, and Olara Otunnu as well, whom that room had not named. It then asks all fifteen delegations which of the three who remain — Grynspan, Rodrigues Birkett, Espinosa — wins their vote.

Every delegation entered its own pair, and the answer is lopsided: 10 of the fifteen put Rebeca Grynspan first, 4 put Carolyn Rodrigues Birkett first, and China alone puts María Fernanda Espinosa first. Ten is above the nine votes a real recommendation needs. The room did not treat it that way: the personas insisted on recording a preference and a vote intention as two different things, and agreed that a name could only be put forward with nine affirmative intentions and no permanent member against. On that test they named nobody.

The more interesting number is the second column. Rodrigues Birkett is the second choice of 11 delegations, which together with her four firsts puts her in every single delegation’s top two. That is why the room agreed to test her as the possible consensus candidate rather than the front-runner. China and Russia were the two permanent members that did not place Grynspan in their top two, and China was careful about how it said so.

“I am conditionally willing to support either of my two preferred candidates following satisfactory answers to our common questions; I reserve my recommendation-vote intention on Rebeca Grynspan Mayufis without declaring opposition.”

France shows what the question does to an answer. In the choices room its first choice had been Macky Sall; with Sall struck from the list, Paris moves to Rodrigues Birkett rather than to Grynspan.

“My first choice within the stipulated shortlist is Carolyn Rodrigues Birkett, preserving my surviving preference after Macky Sall’s exclusion; my second is Rebeca Grynspan Mayufis.”

This is the one room that did not finish. All fifteen rows were filled and five of its six points were agreed, but with fifteen delegations each needing a turn, the second round of confirmations was still running when the room hit its twenty-turn limit. The product records it as “turn limit reached”, not “completed”, and so do we: the table proposal itself was never re-confirmed by everyone.

See all fifteen rows and the tally →

Rooms 5 and 6 · the permanent five · 14 and 17 September

Who would be the toughest budget negotiator?

The Organization is being asked to do more with sharply less, so the last two rooms drop the politics and ask a managerial question instead.

Both rooms began by defining the word. “Toughest” was agreed to mean credible savings and reliable financing while the mandated work still gets done — not the largest promised cuts. That definition decides most of what follows.

“I propose that we judge budget toughness by the ability to secure credible savings and reliable financing while preserving mandated delivery, rather than by promises of the largest cuts.”

In the open room, with all eight candidates available, Rafael Grossi is ranked first by 3 of the five — the United States, Russia and France — and Rebeca Grynspan by the other 2. Grynspan is the second choice of the other 3, so she appears somewhere in all five rankings.

That answer sits oddly beside the veto map, where Russia had named Grossi its most likely veto target the same evening. The two rooms are answering two different questions: who would drive the hardest bargain on money, and whom a capital could accept in the job. A tough negotiator and an electable candidate need not be the same person, and neither room claimed they were.

“Grossi appears the toughest advocate of cuts; that does not yet establish that he would negotiate the strongest budget settlement.”

So the question was put again on 17 September with the likely-vetoed candidates set aside and only the three finalists on the table. The answer is unanimous: all 5 permanent members rank Grynspan first for budget toughness, Rodrigues Birkett a close second for coalition leverage and scrutiny of delivery losses, Espinosa third. Unanimity here is cheaper than it looks: the reasons are nearly identical in every capital because they come from the same briefing document. This is agreement on a reading of three résumés, not a negotiated deal.

“I place Rodrigues Birkett narrowly second for coalition leverage and scrutiny of delivery losses, ahead of Espinosa’s substantial but less financially grounded downsizing programme; this is my assessment for this exercise, not a published U.S. position.”

See both rankings →

Method & caveats

What the simulation gets wrong

• One run per question. Each chapter above is a single room on a single day. Where this page says an answer was stable, that comes from other runs of the same question on 14 and 15 September, and it is said explicitly; everywhere else, assume one run is one run.
• The briefing writes the veto answers. Every delegation in the veto room said its entry came from the supplied dossier. That map is closer to a well-organised echo of the document than to an independent finding.
• The model was run cheaply. All six rooms used one model at a low reasoning effort — fast and inexpensive, and more prone to following the most obvious line in the briefing than to arguing with it.
• Personas are not governments. They reason from the public record. They cannot know instructions, side deals, or what a capital will trade later in the race, and their words are not quotations of the ministers they are modelled on.
• The question decides the answer. Removing five candidates — the four the veto room had named, plus Otunnu — turned a room with four different first choices into a room of ten-to-four. That is a property of the question, not a discovery about the Council.
• Counts are not probabilities. Every number here is a count of delegations inside one simulated room. None of them is a share of anything real, and none of them says how the Council will vote.
• What would change these answers. A briefing refreshed after the real third straw poll, and the same questions run many times with the results shown as tallies instead of as one room’s table.

Every figure in these chapters is derived by one small script from a single hand-checked record of the six rooms, and reproduced as a table below. Nothing is typed in by hand twice.

See every table →