Multilateralism in the Global Governance of Artificial Intelligence

Michal Natorski

REVIEW 3 major objections 4 minor 1 references

Multilateralism in the Global Governance of Artificial Intelligence

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This chapter argues that international AI governance is multi-stakeholder in form but state-dominated in substance, with states retaining decisive power over agenda-setting, negotiation, and soft-law implementation.

desk verdict Abstract-only read because the full text is mojibake; what we can see is a coherent, useful conceptual framing with a state-dominance claim that needs case evidence to hold. read the letter →

arxiv 2508.15397 v1 pith:RJ2NONHV submitted 2025-08-21 cs.CY

Michal Natorski This is my paper

classification cs.CY

keywords artificialintelligencegovernancemultilateralismmulti-stakeholderstatehierarchysoftlawinternationalinstitutionsdiscourseanalysisglobal

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The chapter asks how multilateralism is adapting to artificial intelligence and argues that the adaptation is more state-centric than it appears. It identifies three discourse principles that organize AI multilateralism—epochal change, determinism, and dialectical understanding—and shows that both existing cooperation frameworks and new AI-only bodies operate in the shadow of the state hierarchy. Although non-state actors such as companies and civil society join these forums, the chapter contends that states keep the decisive role in agenda-setting, negotiation, and implementation of non-binding commitments. If true, the multi-stakeholder appearance of AI governance should not be mistaken for a diffusion of authority; formal decision power still sits with states.

What carries the argument

The analysis is carried by two mechanisms. First, a discursive frame made of three principles—epochal change, determinism, and dialectical understanding—organizes how AI is talked about in multilateral settings. Second, the 'shadow of state hierarchy' mechanism describes how existing cooperation frameworks and new AI-only bodies keep authoritative decision points in state hands even when they formally include companies, experts, and civil society. These mechanisms together explain why multi-stakeholder participation and state dominance coexist.

What would settle it

Compile an exhaustive inventory of AI governance bodies—intergovernmental forums, multi-stakeholder initiatives, and technical standards organizations—and code who has final authority at each stage: agenda-setting, negotiation, and implementation. If even a sizable minority of consequential venues grants formal decision power to non-state actors, the claim that states preserve decisive competence fails in its current general form. A second check is textual: apply the three discourse principles to a random sample of AI governance documents; if they do not structure the debates, the shared-disco

Watch

Extended reading notes

Core claim

The core claim is that AI multilateralism is multi-stakeholder in composition but statist in hierarchy. The chapter distinguishes two pathways of adaptation—folding AI issues into existing cooperation frameworks and creating new ad hoc frameworks devoted exclusively to AI—and argues that both reproduce a 'shadow of state hierarchy' over non-state participants. Its account rests on three generalized principles said to structure the discourse: epochal change (AI as a historic rupture), determinism (AI development as driven by technological forces), and dialectical understanding (tensions and contradictions between promises and risks). On this view, states preserve competence as decisive decisi

Load-bearing premise

The load-bearing premise is that the frameworks examined are representative of AI multilateralism as a whole; if the sample tilts toward state-centric bodies, the conclusion would follow from the selection rather than from the phenomenon.

Editorial extensions

If this is right

AI governance researchers and policymakers should treat multistakeholder participation as consultation, not co-decision: final authority over agenda, text, and implementation stays with states.
Because commitments are soft law, compliance depends on state goodwill; private-sector promises inside these forums carry weight only insofar as states endorse them.
Creating new AI-only bodies does not by itself shift power; the same state hierarchy reappears unless formal decision rights are allocated differently.
The three discourse principles can be used to read any AI governance document and to predict which policy options appear legitimate and which are excluded.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural test of the state-hierarchy claim is to map decision rights in industry-led technical standardization bodies, where firms often hold formal voting power; if those bodies set binding standards without state sign-off, the 'shadow' may cover intergovernmental soft law but not the full governance landscape.
The three discourse principles could be operationalized as a coding scheme; a plausible extension would predict that industry venues emphasize determinism and epochal change more than dialectical understanding, while state venues balance all three.
If the claim is right, shifts in AI governance outcomes will track shifts in state interests rather than shifts in corporate or civil-society participation; that is a testable implication for case studies of specific agreements.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

What you need to know up front: the file we received is corrupted beyond the first page, so this is strictly an abstract-only assessment. The abstract is well written and the argument is coherent, but none of the empirical claims can be verified from what we have.

The genuine novelty, such as it is, is the three-principles vocabulary—epochal change, determinism, dialectical understanding—offered as a way to organize AI multilateralism. That is a useful conceptual move for a young subfield, and the abstract's core thesis is a clear one: AI governance is multi-stakeholder in form but state-centered in substance, operating 'in the shadow of the state hierarchy.' The shadow-of-hierarchy idea is familiar from European governance literature, so the contribution is the application to AI, not the concept itself. Still, as a synthesis, the chapter appears to give researchers a compact way to talk about the field.

The soft spot is exactly what the stress-test note identifies: the case selection is invisible. The abstract names no frameworks, no sampling rule, no number of cases, and no evidence. A claim that states remain decisive in agenda-setting, negotiation, and implementation is a generalization over all of AI multilateralism, and a skeptical reader cannot tell whether the corpus is representative. The concrete worry about technical standardization bodies—ISO/IEC JTC 1/SC 42, IEEE, industry consortia where non-state actors hold formal decision rights—is legitimate. If the author excludes or marginalizes those venues, the state-dominance conclusion may follow from the sample rather than the phenomenon. I want to be fair, though: we cannot see whether the full chapter addresses this, because the text is unreadable. The same goes for the three principles; they may be grounded in careful discourse analysis, but no coding procedure is visible in the abstract.

This is a paper for AI governance researchers and IR scholars who want a conceptual frame rather than a new empirical dataset. It deserves a serious referee, but only after the author supplies a clean, complete PDF. Desk rejection on the basis of a corrupted file would be unfair, and the abstract suggests the full chapter could be a legitimate contribution. I would not cite it until I can read the full text, but I would put it on the reading list for the next workshop.

Recommendation: send to peer review with a request for a corrected manuscript.

Referee Report

3 major / 4 minor

Summary. This chapter proposes an analysis of how international multilateralism is responding to artificial intelligence. The abstract makes two main claims: first, that AI multilateralism is organized by three generalized principles (epochal change, determinism, dialectical understanding); second, that although AI governance appears multi-stakeholder, states remain the decisive decision-makers in agenda-setting, negotiation, and implementation of soft law international commitments, so that AI multilateralism develops in the shadow of the state hierarchy. It further claims that the field is structured by the absorption of AI into existing cooperation frameworks and by the creation of new AI-only frameworks. The submitted arXiv full text is, however, largely unreadable because of an encoding corruption; apart from the abstract and fragments of section headings, only mojibake is visible. Consequently, the empirical basis, case selection, and discourse-analytic method behind the stated conclusions cannot be inspected.

Significance. If the central thesis is substantiated, the chapter would make a useful intervention in a genuine debate: whether global AI governance is genuinely multistakeholder or remains state-centric in substance. The claim is clear and falsifiable in principle: one could test it against a broad sample of AI governance venues, including technical standardization bodies, industry consortia, and multistakeholder initiatives. The three proposed organizing principles could also serve as a heuristic framework for later empirical work. The paper offers no machine-checkable proofs or reproducible code, but for a qualitative IR chapter that is not expected; what is required is transparent qualitative evidence. Unfortunately, the current submission does not provide that evidence in any inspectable form, so the significance is conditional on a substantially revised version being made available.

major comments (3)

[Abstract] The central generalization about 'AI multilateralism' rests on an unstated sampling premise. The abstract names no frameworks, no selection rule, and no number of cases, and its binary distinction between existing cooperation frameworks and new AI-only frameworks is not a sampling frame. In particular, it is unclear whether technical standardization venues (e.g., ISO/IEC JTC 1/SC 42, IEEE) or industry-led multistakeholder bodies, where non-state actors hold formal decision rights, are part of the analysis. If such venues are excluded, the state-hierarchy conclusion may follow from the case selection rather than from the phenomenon. The manuscript should specify the full corpus, the inclusion/exclusion criteria, and a representativeness argument.
[Full text (submitted version)] The body of the submitted paper is not legible: it consists of corrupted character sequences, so no method, document analysis, coding protocol, or case-level evidence can be checked. This is load-bearing because the abstract's empirical claims cannot be evaluated without the underlying qualitative analysis. The authors need to resubmit a readable text, and the methods section must explain how documents were selected, how 'decisive decision-maker' was operationalized, and how the three principles were derived from the discourse.
[Three generalized principles] The abstract presents 'epochal change, determinism, and dialectical understanding' as generalized principles of AI multilateralism, but nowhere in the available text is there a coding procedure, source quotations, or an account of how these categories emerged from the material rather than being imposed on it. This creates a circularity risk: if the same discourse is used both to extract the principles and to confirm them, the principles are not independently evidenced. The chapter needs to show the analytical protocol, give examples of coded passages, and address the question of inter-coder reliability or at least analytic transparency.

minor comments (4)

[Abstract] The phrase 'states preserve the competence as decisive decision-makers' is awkward; 'competence' should probably be 'authority' or 'competency' in this context. The sentence should be rewritten for clarity.
[Abstract] The term 'general-purpose technology' is used without citation; if not cited elsewhere in the readable portion, a reference to the standard GPT literature (e.g., Bresnahan and Trajtenberg) should be added.
[Full text] The reference list, if present, is not recoverable from the corrupted text. A complete and correctly rendered bibliography is required.
[Section 1 / framing] The manuscript should define its scope with respect to 'multilateralism.' Does the term include regional bodies, minilateral forums, technical standardization organizations, and industry consortia, or only global intergovernmental organizations? The current abstract implies a global scope without defining boundaries.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found; the chapter's claims are empirical/interpretive generalizations, not results forced by construction or self-citation.

full rationale

The paper is an interpretive, qualitative study of multilateral AI governance rather than a derivation chain. Its central claims—that AI multilateralism operates in the shadow of state hierarchy and that its discourse is organized by principles of epochal change, determinism, and dialectical understanding—are presented as empirical generalizations about cooperation frameworks, not as outputs of equations fitted to inputs. No fitted parameter is relabeled as a prediction, no definition is shown to presuppose the conclusion, and no load-bearing self-citation or imported uniqueness theorem appears in the abstract or readable fragments. The potential concern that the case selection may be unrepresentative is a sampling/validity issue, not circularity: even if the selected frameworks are state-centric, the conclusion would then be an overgeneralization from evidence, not a tautology. Similarly, the risk that the three 'generalized principles' were imposed on rather than found in the discourse is an interpretive-method concern; without a coding procedure to inspect, it cannot be established that the explanation is equivalent to its input by construction. Because no specific reduction can be quoted, the honest finding is no significant circularity.

Assumptions & free parameters 0 free parameters · 3 assumptions · 1 invented entities

No quantitative claims are made, so there are no fitted numbers. The load-bearing burdens are interpretive and sampling-related: representativeness of the frameworks examined, grounding of the three principles in the discourse, and the assumption that soft-law venues are where AI governance power is decided. None of these is justified in the abstract.

assumptions (3)

domain assumption The set of analyzed international frameworks is representative of AI multilateralism as a whole.
The abstract describes the analysis as covering integration into existing cooperation frameworks plus new AI-only frameworks, but names no frameworks, no selection rule, and no number of cases. If the corpus favors state-centric bodies, the state-hierarchy conclusion follows from the sample.
domain assumption The three generalized principles (epochal change, determinism, dialectical understanding) are present in AI governance discourse rather than imposed by the analyst.
The abstract asserts the chapter 'distinguishes' these principles but shows no coding procedure or textual evidence. This is the standard interpretive-coding risk and it is load-bearing for the first half of the chapter.
domain assumption Soft-law international commitments are the decisive venue of AI governance.
The conclusion that states are 'decisive decision-makers' is tied to agenda-setting, negotiation, and implementation of soft law. The chapter implicitly assumes that legally binding instruments and technical standard-setting bodies are not where the real decisions happen.

invented entities (1)

The three generalized principles of AI multilateralism (epochal change, determinism, dialectical understanding)
purpose: Analytical categories used to organize the description of AI governance discourse across international frameworks.
These are the author's interpretive constructs. They carry no falsifiable handle outside the chapter, so they function as an invented conceptual ledger rather than an empirically anchored discovery.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Multilateralism in the Global Governance of Artificial Intelligence." pith.science (2026). https://pith.science/paper/RJ2NONHV

@misc{pith2026250815397,
  author       = {Pith},
  title        = {Pith review of: Multilateralism in the Global Governance of Artificial Intelligence},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RJ2NONHV}},
  note         = {Machine review of arXiv:2508.15397}
}

read the original abstract

This chapter inquires how international multilateralism addresses the emergence of the general-purpose technology of Artificial Intelligence. In more detail, it analyses two key features of AI multilateralism: its generalized principles and the coordination of state relations in the realm of AI. Firstly, it distinguishes the generalized principles of AI multilateralism of epochal change, determinism, and dialectical understanding. In the second place, the adaptation of multilateralism to AI led to the integration of AI issues into the agendas of existing cooperation frameworks and the creation of new ad hoc frameworks focusing exclusively on AI issues. In both cases, AI multilateralism develops in the shadow of the state hierarchy in relations with other AI stakeholders. While AI multilateralism is multi-stakeholder, and the hierarchy between state and non-state actors may seem blurred, states preserve the competence as decisive decision-makers in agenda-setting, negotiation, and implementation of soft law international commitments.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

1 extracted references · 1 canonical work pages

[1]

�� ...

work page Pith review arXiv 2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

�� ...

work page Pith review arXiv 2025