Second Table, Same Ceiling: Two Systems of Legitimation, One Structural Limit

Russian-language version available

Contents

Shanghai, 16 July 2026. On the eve of the World AI Conference, the Minister of Digital Technologies of Uzbekistan signs an agreement, and his country becomes a founding member of a new global organization for artificial intelligence. Nine days earlier, and five weeks before that, two other tables had been set on the other side of the world. The world now has two international tables for AI governance, not one. This piece does not assess the decision any government took. It examines the structure of the tables themselves, a structure that applies in the same way to every state of similar material position, whichever table it sits at. The question is narrower than the headlines, and older than the event: what changes when the world no longer gathers around a single institutional table, and what does not.

I. A Decade of Layering, Compressed Into a Month

The signing in Shanghai reads as a sudden event only if the years before it are removed from view. They should not be. What follows is not a list of organizations. It is a sequence of institutional layers, each settling on top of the last.

The direction was set in 2023, when China issued its Global AI Governance Initiative, a declaration organized around sovereignty, development, and openness rather than around a shared political standard for entry [1]. In July 2025, at the World AI Conference, Premier Li Qiang announced that this declaration would be given an institutional body, the World Artificial Intelligence Cooperation Organization, to be seated in Shanghai, alongside a thirteen-point action plan [2]. President Xi Jinping restated the proposal at the APEC leaders’ meeting that October. In December 2025, from the other direction, the United States launched Pax Silica, described by its own architect as a purpose-built coalition of trusted partners for the AI economy, explicitly set against the language of digital sovereignty and explicitly distinguished from broader forums such as the G7 and G20 [3].

Then came the compression. In late June 2026, the second Pax Silica Summit in Washington produced the AI Opportunity Statement, aligning a group of partner economies on a pro-innovation, supply-chain-security approach to AI [4]. On 7 July, in Geneva, the United Nations closed the first Global Dialogue on AI Governance, a universal forum that produced, by its own design, nothing enforceable, for reasons examined earlier in this series [5]. Nine days after Geneva, the WAICO agreement was signed in Shanghai by twenty-nine founding states, with the United Nations Secretary-General in attendance and the Chinese foreign minister signing on behalf of his government [6].

Set in sequence, WAICO is not a rupture. It is the most recent layer in a field that has been stratifying for years, and the month of July 2026 simply made the stratification visible from a single vantage point. The interesting object is not the newest institution. It is the pattern the institutions now form.

II. Two Systems of Legitimation

It is tempting to describe what has emerged as two lists of countries, or two organizations, or two great powers. All three descriptions are true and all three are too small. What has actually formed is two competing systems for structuring international legitimacy in the governance of artificial intelligence: two different answers to the question of where the authority to set the rules is now supposed to come from.

The systems are built on opposite logics of belonging. The academic mapping of the field, produced this year from within Tsinghua University, codes fifteen existing AI governance bodies and finds that WAICO was designed to occupy a position no constituted institution then held: membership presented as open to any sovereign state, with no values test at the door, and an agenda built around development and the closing of the global capability divide [7]. The Western and like-minded bodies sit at the opposite pole. Pax Silica is not offered to whoever is sovereign. It is offered to trusted partners, and its stated purpose is to align the AI supply chains of its members with one another and with the United States [3][4]. One system asks a state to be sovereign. The other asks it to be the right kind of partner. That is not a difference of tone. It is a difference in the source of legitimacy each system claims to draw on, one appealing to universal inclusion, the other to shared commitment.

A definition is worth fixing here, so the word does not drift. In this analysis, legitimacy means not moral authority but accepted authority to coordinate collective action, the standing that lets a governance framework claim a role in setting rules others are expected to observe.

And here the more important symmetry appears, the one the competition obscures. Neither system, on the documentation available in mid-July 2026, carries the capacity that the first table was already found to lack. The Geneva Dialogue produced no independent verification mandate because none was structurally available to it. WAICO, as of its founding, has published no charter, no secretariat, no budget, and no decision rules; its own participants describe the priorities of the organization as still to be determined and its members as future member states [6][8]. The establishment agreement does provide that a governing council will set annual financial contributions on a scale modeled loosely on the Universal Postal Union, so some machinery is named [9].

But a contribution formula is not a verification power. On the measure of institutional formalization used in the Tsinghua coding, WAICO scores lowest of the fifteen bodies examined [7]. Pax Silica, for its part, is a coalition for supply-chain alignment, not a body that inspects or halts the systems its members deploy.

So the field now has a second table. What it does not have, at either table, is the ability to independently verify the systems that both tables exist to govern. This is the same ceiling documented after Geneva, reproduced rather than removed. Two tables, one ceiling.

III. The Registries Barely Touch

If two systems of legitimation are forming, the sharpest evidence is not who joined either one. It is how little they overlap.

The point can be checked directly, which matters, because a claim used as an argument should carry its own method. The State Department publishes the signatories of the AI Opportunity Statement, thirty-five in all. It separately publishes the signatories of the Pax Silica Declaration, a distinct document, twenty-three states and the European Union. China’s official channels and Uzbekistan’s own government confirm the twenty-nine founding members of WAICO. Cross-referencing the WAICO founders against both United States documents yields a single name.

Body Lead Entry logic Independent verification mandate Central Asian members
WAICO founding members (29) China broadly inclusive Not documented Kazakhstan, Kyrgyzstan, Tajikistan, Uzbekistan
AI Opportunity Statement (35) United States trusted partners Not documented Kazakhstan
Pax Silica Declaration (23 states + EU) United States trusted partners Not documented Kazakhstan

Of the twenty-nine WAICO founders, exactly one, Kazakhstan, also appears on either United States list. At the level of membership, the two systems of legitimation are almost perfectly disjoint.

The near-absence of overlap is the structural signal. Kazakhstan is the exception that measures how narrow the bridge is, not the center of the story. And Central Asia turns out to be a clean instrument for reading the pattern, because the region distributes itself across the divide: one state on both sides, three states, including Uzbekistan, present at the eastern table and absent from the western documents.

Uzbekistan’s position is worth stating with precision, because precision is also protection. Begin with how the country describes its own participation. Its government frames its WAICO membership in the language of international cooperation, technology exchange, and its long-term strategy to become a regional leader in artificial intelligence, a technocratic framing, not a geopolitical one [10]. On the documented facts, the country is a founding member of WAICO, its agreement signed by the Minister of Digital Technologies, Sherzod Shermatov [10]. And on the western side, the country does not appear among the signatories of either the AI Opportunity Statement or the Pax Silica Declaration [4].

What is not documented is any account, from Washington or from Tashkent, of why Uzbekistan is absent from the western documents. The United States publishes no membership criteria and no application procedure for these instruments; states are described as joining by signature, in a process that appears curated rather than open [4][11]. Absence from a curated list is not, on the available record, a choice by the absent state. It may equally be the result of a door whose opening is decided by the host. The record does not say which, and neither, therefore, will this analysis.

One further fact belongs beside the others, and it cuts against any tidy story of alignment. According to the Uzbek outlet UzDaily, in early June 2026, five weeks before the Shanghai signing, Minister Shermatov received a United States Deputy Secretary of Commerce in Tashkent for talks on artificial intelligence, data centers, computing capacity, and cybersecurity [12]. That meeting was not documented on the government portal, nor confirmed in public United States government sources at the time of writing, and it should be read only at the weight its single non-state source can bear. But at that weight it still says something. A state can be building an AI relationship with Washington and founding an organization in Shanghai in the same season. The eastern table and the bilateral conversation with the West are not a contradiction in Tashkent’s conduct. They are two projections of one position.

IV. Representation Acquired, Control Unchanged

The appearance of a second table changes the geography of international legitimacy. It does not change the geography of institutional capability. To see why, return to the position this series has called the structural situation of the Global South, defined not by geography but by place in the AI production stack: consumption without design, subordination without representation, responsibility without control, sovereignty without material power [13].

Read against those four dimensions, founding membership in WAICO changes exactly one, and only in part. It touches representation. A seat at a founding table is a form of voice, and it is not nothing; a state that helped establish an organization is, formally, a maker of that organization rather than a recipient of its outputs. On the second dimension, then, the signature registers. On the other three, it does not reach. It does not add fabrication capacity, or frontier compute, or an independent means of verifying the systems the organization exists to govern. The distribution of the material substrate is unchanged by the ceremony. A state may now sit where the rules will be discussed without gaining any greater capacity to make those rules answer to it, because the capacity in question was never at the table to be distributed.

This is why the competition between the two systems of legitimation matters less, for a state in Uzbekistan’s position, than it appears. The choice a state is offered, in the framing of both camps, is presented as a choice of whose rules to help write. The choice actually available is thinner. It is a choice of which legitimating table to sit at, while the material and verification parameters that determine governance outcomes remain, at both tables, held elsewhere. What both tables can hand out is a seat. What neither was ever able to hand out is the substrate beneath it.

Implications

Three implications follow, and they are best kept separate by the confidence each deserves.

What is documented: the field of AI governance now has two competing institutional poles rather than one, each with its own logic of membership, and their memberships are almost entirely disjoint. Neither pole, on its founding documentation, carries an independent verification mandate over the systems it governs. Uzbekistan is a founding member of the eastern institution and absent from the western documents, and its government frames its participation in cooperative, developmental terms.

What is not confirmed: why any particular state is absent from the western documents; whether WAICO’s membership will remain as configured once its founding period closes; whether the reported Tashkent–Washington bilateral talks reflect a durable track or a single meeting.

The analytical inference, offered as inference and not as fact: the appearance of a second table does not relieve the structural limit the first table was already found to have. It relocates the limit into a second form. For states that consume frontier AI rather than design it, the multiplication of tables widens the menu of legitimating memberships without widening the narrow band of actual control, and engagement at either table is therefore best understood as movement within an inherited architecture rather than a redesign of it. More tables, the same ceiling.

Signals to Watch

Several markers will show whether this reading holds.

  • Whether WAICO, as it moves from signed agreement toward operating body, is given a charter, a secretariat, a budget, and decision rules, and in particular whether any organ is granted a verification or audit mandate over member systems rather than a coordinating role;
  • Whether the founding period closes on 31 July 2026, as reported by member-state sources, or remains open beyond it, and whether the final roster differs from the twenty-nine, and whether any state other than Kazakhstan comes to sit at both tables;
  • Whether the material commitments that accompany membership, the training programs and tool access announced in Shanghai, arrive on terms set by the host or negotiable by the members;
  • and whether the next session of the United Nations Global Dialogue, in New York in 2027, engages the existence of a parallel pole or proceeds as though the field still had one center.

The Questions That Remain Open

Two questions this piece cannot resolve. First, whether two competing registries can generate the distributed capacity for verification that a single universal forum could not, or whether the division simply reproduces, twice over, the same distance between governance claim and operational effect. Second, whether formal representation in an expanding institution can, over time, convert into genuine influence over the material and verification parameters that decide governance outcomes, or whether a seat at a legitimating table is precisely the thing that is offered in place of that influence.

Sources & Notes

[1] Ministry of Foreign Affairs of the People’s Republic of China. Global AI Governance Initiative. 2023. mfa.gov.cn

[2] The State Council of the People’s Republic of China. Chinese premier calls for early formation of global AI governance framework. 26 July 2025. gov.cn

[3] U.S. Department of State. Pax Silica and associated releases; Under Secretary Jacob Helberg’s framing of the initiative as a purpose-built partnership for the AI economy, distinct from the G7 and G20. state.gov

[4] U.S. Department of State. AI Opportunity Statement and Outcomes of the Second Pax Silica Summit. June 2026. Signatory lists for the AI Opportunity Statement (35) and the Pax Silica Declaration (23 states and the European Union). state.gov

[5] Khodjaev, Oybek. The Verification Ceiling: What Geneva Confirmed About Global AI Governance. 9 July 2026. okhodjaev.com

[6] Xinhua / State Council of the People’s Republic of China. 29 countries sign agreement on establishing World AI Cooperation Organization (WAICO). 16–17 July 2026. gov.cn

[7] Guey, W., Bougault, P., Zhang, W., de Moura, V. D., and Gomes, J. O. World Artificial Intelligence Cooperation Organization (WAICO): Mapping an Emerging Institution in the Global AI Governance Regime Complex. arXiv:2606.23860, June 2026. The paper codes fifteen AI governance instruments on three variable groups — membership, formalization, and normative orientation — and reports WAICO’s position on the membership-and-orientation axes together with its formalization score, the lowest in the sample. arxiv.org

[8] Ministry of Digital Technologies of the Republic of Uzbekistan. Minister of Digital Technologies Participates in Preparatory Session on the Establishment of WAICO. 18 July 2026. The ministry’s own account refers to the organization’s future priorities and to future member states. gov.uz

[9] Global New Light of Myanmar. Proposals submitted to Pyidaungsu Hluttaw to extend validity of Ordinance 1/2026, join WAICO. July 2026. In a statement to Myanmar’s parliament, the Deputy Minister for Digital Development and Communications reported that the WAICO signing period runs from 1 May to 31 July 2026, and that a council will set annual member contributions on a scale drawing on the Universal Postal Union model. This is a member-state parliamentary account, corroborated by Indonesian reporting citing a ministerial statement; it is not, at the time of writing, confirmed by a WAICO or Chinese primary document. gnlm.com.mm

[10] Ministry of Digital Technologies of the Republic of Uzbekistan. Uzbekistan Becomes a Founding Member of the World Artificial Intelligence Cooperation Organization (WAICO). 16 July 2026. Signature by Minister Sherzod Shermatov; official framing of participation. gov.uz

[11] On the absence of any published accession criteria or application procedure for the Pax Silica instruments: assessment based on the public State Department pages cited at [3] and [4]; no such procedure is documented there. Any characterization of the process as invitation-based is inference, not a documented rule.

[12] UzDaily. Uzbekistan, US Eye AI and Cybersecurity Partnership in Tech Talks. 9 June 2026. A non-state Uzbek outlet; the meeting was not documented on the government portal and was not confirmed in public United States government sources at the time of writing. The source names a United States Deputy Secretary of Commerce as the delegation lead; that individual detail has not been independently verified against official United States records and is not relied upon in the body of this analysis. uzdaily.uz

[13] The four-dimensional account of the Global South as a structural position appears in The Colonial Pattern (Essay 5) and The Sovereignty Question (Essay 9) of the Beyond Control series. okhodjaev.com

Full analysis and updated sources: okhodjaev.com/analysis/second-table-same-ceiling

Oybek Khodjaev researches AI governance and advises on institutional readiness in the age of AI, with a particular focus on systemic risk. Drawing on more than 35 years of experience across banking, finance, public administration, and business in Uzbekistan and the CIS, he develops practical approaches to assessing the institutional limits of AI governance. He is the author of the “Beyond Control” essay series, which advances a structural theory of the limits of AI governance.

Published: