Research: IGO Capture — Measuring Leverage at WHO, the UN, Davos, the IMF, the WTO and the AI Bodies

Companion pages: The Health Governance Ratchet, Economic Statecraft, The Capture Mechanism, and the Tradecraft and Capture Leaderboard tools that compute over this data.

The method. Measure, per entity and from each body’s own records, the leverage individual funders and member states hold over the intergovernmental organizations: who pays, who sits, who writes the standards, who can veto. Never a verdict on a faction, never a motive. “Capture” is a label critics apply; this page reports the structure that the label is about, and lets the reader weigh it. The bodies differ, and so do the levers: funding share at WHO, seats at the Human Rights Council, fees and alumni at Davos, a weighted veto at the IMF, an informal back room at the WTO. The point of measuring them first is to apply the same read, with the same discipline, to the bodies that now govern AI.

1. WHO: funding share and earmarking

From WHO’s own funding pages:

• More than 75 percent of WHO’s financing is voluntary, not assessed member dues; assessed contributions cover under 20 percent of the budget (WHO, How WHO is funded).
• Over 70 percent of revenue is earmarked by the donor for a specific programme or country. Earmarking is the hard leverage signal: it turns a donation into a directed budget line.
• The top ten contributors supply 62 percent of all funds (about US$3.65 billion). Foundations account for about 10 percent, led by the Gates Foundation (WHO, current state of financing).

Who is at the top, from WHO’s contributor portal (open.who.int; API data):

A private foundation went from third to first as the largest member state’s contribution fell and then stopped, and GAVI, which the Gates Foundation seeded and lists as its largest grantee, sits third in 2024-25. See the Gates Foundation and WHO profiles.

2026: the largest member state left. The United States’ withdrawal, ordered on January 20, 2025, took effect on January 22, 2026 after the one-year notice period; WHO said the US had not paid its invoiced assessed contributions for 2024-25. Director-General Tedros: “Unfortunately, the reasons cited for the US decision to withdraw from WHO are untrue. The notification of withdrawal makes both the US and the world less safe. We hope the US will return to active participation in WHO in the future” (ANI via Business Standard, Jan 25 2026). With one of its largest contributors gone, the share held by the remaining voluntary and foundation money rises unless member states replace it.

The funding share and the earmarking are WHO’s own facts. Whether they amount to capture of health agenda-setting is the critics’ argument; the case for the arrangement is that earmarked philanthropy pays for programmes member states will not fund.

2. The UN Human Rights Council: seats and standard-setting

The Council’s leverage is not money but who sits and who shapes the standards. It has repeatedly seated states with documented rights-abuse records (China, Cuba, Saudi Arabia, Russia, Venezuela, Iran), which critics say use the seat as a “badge of legitimacy” (International Bar Association). UN Watch, an advocacy group, assessed one election as having only 4 of 16 candidates qualified on their records, and documents that China was placed in 2020 on the five-member Consultative Group that helps select the Council’s independent experts and special rapporteurs: a move from member under review to standard-setter (UN Watch database). The United States withdrew in 2018 citing membership and structural bias.

It left again in 2025. Executive Order 14199 of February 4, 2025 ended US participation in the Council, ended funding to UNRWA, and ordered a review of UNESCO (Federal Register, Feb 10 2025). On July 22, 2025 the administration announced the US would leave UNESCO effective December 31, 2026, the State Department saying membership is not in the US “national interest” and the White House calling UNESCO a supporter of “woke, divisive cultural and social causes” (CNN via KTVZ, Jul 22 2025). Exit is the bluntest per-entity lever there is: the member takes its dues, its vote and its staff, and the body’s remaining donors and voting blocs gain share.

3. The World Economic Forum: fees and the alumni pipeline

• The price ladder. About 100 “Strategic Partner” companies pay up to about US$758,000 a year; some spend about US$1 million on Davos sponsorship (WEF pricing; Yahoo Finance, “Inside Davos’ pricey power club”).
• Young Global Leaders. Schwab’s programme (launched 2004) counts about 1,400 members and alumni across 120 nationalities (InfluenceWatch), including sitting and former heads of government and ministers: Macron, Ardern, Marin, Spahn (The Spectator, “Is Klaus Schwab running France?”). The measurable metric is the count of heads of government and cabinet members who are alumni.
• Schwab’s own words. At the Harvard Institute of Politics forum on September 20, 2017, moderated by David Gergen (Harvard Crimson), he described how “we penetrate the cabinets” of governments through the programme and claimed that half of Justin Trudeau’s cabinet were Young Global Leaders (the clip, on X). The Spectator Australia checked the roster: “This may have been an exaggeration: five were, and although Schwab lay claim to nurturing Trudeau himself, it isn’t clear whether he was actually a ‘Young Leader’” (The Spectator Australia, Jan 2024). The fee ladder, the roster and his own boast carry the point; the “Great Reset” story that the Forum secretly runs governments belongs to its promoters and is not adopted here.
• The founder left. Schwab resigned as chair in April 2025. In August the board said a law-firm review found no “material wrongdoing” by him, and named BlackRock’s Larry Fink and Roche vice chairman André Hoffmann interim co-chairs (Bloomberg via Swissinfo, Aug 15 2025). The fee ladder and the alumni programme are unchanged; the chair now sits with the chief executive of the world’s largest asset manager. See World Economic Forum and Larry Fink.

4. The IMF and World Bank: a weighted veto and conditionality

• The US veto. The IMF’s most important decisions (admitting members, quota increases, SDR allocations, amending the Articles) need an 85 percent supermajority. The United States holds about 16.5 percent of the vote, above the 15 percent blocking line, so it alone can veto every one of them; no other single member can (IMF, How the IMF Makes Decisions; IMF quotas and voting power).
• The top jobs are allocated. By unwritten convention the US names the World Bank president and Europe names the IMF managing director.
• Scale. The US had contributed about US$161 billion to the IMF as of March 2022. An academic treatment argues that because the US “is single-handedly able to block major changes,” the institution “is vulnerable to instrumentalization in the pursuit of US interests” (Frontiers in Political Science, “Empire by design,” 2026).
• Conditionality. Fund and Bank lending carries structural conditions: deregulation, privatization, austerity. The IMF’s own 2018 review found the number of structural conditions rising despite pledges to streamline them (Bretton Woods Project), and an 81-country study of 1986-2016 programmes found the reforms “trapped more individuals in cycles of poverty… fostering dependency” (Harvard International Review). “Neocolonial” and “empire” are those authors’ words. The strongest case for conditionality is that a lender of last resort that lends without conditions finances the policies that produced the crisis, and is not repaid. See the IMF profile.

The leverage here is the hardest number in the set: a structural vote share, plus a count of conditions per programme.

5. The WTO: the Green Room

The WTO decides by consensus, but the operative mechanism is the informal “Green Room”: a small, self-selected group of developed states plus a few large developing economies (Brazil, India, China, South Africa, Mexico and a handful of others) settles the divisive issues and forwards the result to the full membership; most smaller developing countries “stay out for lack of adequate resources or capabilities” (Peterson Institute, “Decision-Making in the WTO”). Critics argue the process manufactures a “false consensus” through bilateral pressure and reflects “the political and economic power of its most influential members” (CIEL and allied NGOs). The metric is presence in the room.

6. The AI-governance bodies: the same signature

Run the same three reads (funding, revolving door, standard-setting) on the bodies that govern AI.

Funding and self-governance. The Frontier Model Forum was founded in July 2023 by Anthropic, Google, Microsoft and OpenAI, which set its charter, governance and funding; it develops “safety best practices and standards” for frontier models and draws on “the technical and operational expertise of its member companies” (Microsoft, Jul 26 2023; Frontier Model Forum). The governed fund and run the body that writes the rules: a funder share of about 100 percent industry. See the Frontier Model Forum profile.

The pattern reached the White House on September 29, 2026: a voluntary, “morally binding” one-page “Joint Commitment on Frontier Responsibilities,” signed by the President and the heads of Google, Anthropic, Meta, OpenAI, X and NVIDIA, under which each company picks an external auditor, reports to its own board committee, and polices its own controls, with codification left for “over time” (Reuters via The Star, Sep 29 2026).

The revolving door. The US AI Safety Institute (now the Center for AI Standards and Innovation, CAISI) staffed up from the lab-adjacent evaluation world (METR, Apollo Research). A bipartisan House Science Committee letter in December 2024 criticized NIST for a lack of transparency and no competitive process over a planned grant to RAND, and NIST staff reportedly objected to an expected appointment from the effective altruism milieu (VentureBeat; Centre for Future Generations, the AI Safety Institute network). The sources note it cuts both ways: fast capacity-building and an independence risk. In 2026 the institute’s leadership stayed unsettled: Chris Fall, appointed director in April, resigned after three months, with NIST Director Arvind Raman acting (WJLA/TNND, citing Axios). And on September 9, 2026 OpenAI put Paul Christiano, a CAISI senior technical adviser and the institute’s former head of AI safety, on the OpenAI Foundation board and its Safety and Security Committee, noting he “will recuse himself from all OpenAI-related matters as well as all model evaluations” (OpenAI). The fuller map is Watching the Watchers.

Standard-setting and access. On August 29, 2024 the US institute signed agreements with Anthropic and OpenAI for pre- and post-deployment model access (NIST). The body that judges the models needs the labs’ cooperation to see them, while the labs, through the Forum, define what “safety” means.

The intelligence wire. Paul Nakasone, NSA director and commander of US Cyber Command (2018-2024), joined OpenAI’s board in 2024 and its safety and security committee (AP via ABC News, Jun 13 2024). It is a documented board seat, not a directed operation, but it is the sharpest point where the intelligence and AI-governance graphs touch. See Paul Nakasone.

7. The EU AI Office and the standard writers

The office. The EU AI Office, inside the Commission, enforces the AI Act (European Commission); some MEPs had proposed it as an independent entity, and “it was integrated into the Commission in a compromise” made by rapporteur Dragoș Tudorache for regulatory coherence (EU AI Act Newsletter #50, Future of Life Institute, citing Euractiv). Corporate Europe Observatory, an advocacy group, reported that companies developing AI models “enjoyed privileged access to drafting the Code of Practice” while civil-society organisations “were sidelined” (CEO, Jul 2025).

The pause campaign, and the pause. The Computer & Communications Industry Association (members include Apple, Meta and Amazon) and the “EU AI Champions Initiative” pushed a two-year “stop-the-clock” pause and a “simplification” of the Act, with Apple, Google and Meta campaigning publicly against it (CIO). In 2026 they got most of it: the “AI Omnibus” amending the Act was formally adopted and took effect in late July, moving the core high-risk obligations from August 2026 to 2 December 2027 for stand-alone systems and 2 August 2028 for AI embedded in regulated products, and extending the transparency grace period (Addleshaw Goddard, Jul 2026).

The standard is the law. The Act delegates most technical requirements to harmonised standards written by CEN-CENELEC JTC 21 (established June 2021; about 300 experts from more than 20 countries, five working groups). Compliance with the standard grants a presumption of conformity, so whoever writes the standard writes the binding rule (European Commission, standardisation of the AI Act). The Commission itself describes standardization as “typically seen as a matter of principal interest to industry and technical experts,” and JTC 21 had to stand up a Task Group on Inclusiveness to “bring new stakeholders” and “broaden civil-society input” (AI Standards Hub); Marta Cantero Gamito calls the arrangement “delegated technical governance” (Cantero Gamito, Nov 2025).

8. The counter-case: NIST’s AI Risk Management Framework

Not every body fits. NIST’s AI Risk Management Framework (1.0, January 26, 2023) was developed over 18 months through an open process with more than 240 contributing organizations from industry, academia, civil society and government: a request for information, several public drafts and workshops (NIST, AI RMF development; NIST AI RMF). Industry took part heavily (IBM, AWS, the Business Software Alliance, the US Chamber of Commerce), but so did everyone else. On this record the US technical-standards process is less captured than the EU’s harmonised-standards process, and saying so is what keeps the rest of the read credible. The governance body (the institute) and the framework’s drafting are separate questions.

9. Five channels, one pattern

Five bodies, five different mechanisms, one recurring pattern: the output is steered by a concentrated set of the interested. Where AI governance now sits, that concentrated set is the regulated industry itself. The same read applied to WHO turned up a foundation at the top of the donor table; applied to AI, it turns up the labs at the top of every table. That is what makes “capture” a checkable, receipt-backed structure rather than a slogan, and the NIST counter-case is what keeps it honest.

Related

• The Health Governance Ratchet · Economic Statecraft · Nation-State Ratchets
• The Capture Mechanism · The Funding Ratchet
• Watching the Watchers · The EA to AI-Safety Funding Pipeline · Lawfare
• Tradecraft · Capture Leaderboard · AI Governance Tracker
• Profiles: Gates Foundation · WHO · World Economic Forum · Larry Fink · IMF · Frontier Model Forum · Paul Nakasone