“AI is advancing at a runway speed. The question is no longer whether AI will transform our world – it already is. The question is whether we will govern this transformation together – or let it govern us.”
—António Guterres, Secretary-General of the United Nations
Introduction
On 6–7 July 2026, the United Nations (‘UN’) convened the first annual Global Dialogue on Artificial Intelligence Governance in Geneva (see here and here). The meeting brought together more than 4,200 participants from nearly 170 States, alongside representatives of international organisations, technology companies, civil society, academia and the technical community (see here). Established by consensus through UN General Assembly Resolution 79/325, the Dialogue represents the first universal forum operating under a UN General Assembly mandate in which governments and other stakeholders can deliberate collectively on AI governance. The Dialogue will next reconvene in May 2027 in New York.
Until now, international AI governance has mainly developed through fragmented national and regional initiatives or comparatively exclusive forums such as the G 7, G20, OECD, and World Economic Forum. The Dialogue does not replace these processes. Rather, it provides an intersecting, potentially connective space in which their different regulatory approaches can be compared, contested (and where appropriate) made interoperable.
This post explores how the Dialogue may serve as a multilateral platform for global AI governance. It examines the Dialogue’s institutional and normative aspirations and then identifies the major challenges impairing the success of the Dialogue.
The Dialogue and access to technology
As described by the Finnish delegation, the Dialogue is an “inclusive forum” where “all governments and stakeholders” can engage in “open dialogue”. But dialogues are not concurrent monologues. In the UN context, inclusivity normally involves a minimum shared understanding of problems around which parties can anchor a discussion. So, what are those anchors?
From participants’ statements we can tell what they are not. There is not a single reference to human rights —or “dignity” or “freedom”—in the statement of the Group of 77 and China; nor in the Group of Friends for International Cooperation on AI Capacity-Building’s statement, Türkiye’s or Japan’s statements. In contrast, there are explicit references to human-centric AI “grounded in international human rights law” in the statements of the EU, Australia and Uruguay. Evidently, the need for a global human rights-based approach to AI divides jurisdictions.
A common understanding will not come from basic governance principles either. Although participants frequently called for some coordination around regulatory initiatives or minimal governance, the statements offer very little in terms of a relevant forum to ensure that AI development does not evolve into a risk for peace or security.
However, there is one recurring theme which perfectly fits the UN mandate, is built around its development pillar, and could actually deliver some commitments. That is the theme of universal access to AI technology, alongside governance capability to establish conditions surrounding that access.
One of us participated in a panel on the sidelines of the dialogue on bridging the North-South hardware divide (see generally here). A key panel conclusion was that bridging the divide is possible, but will require persistent engagement, continued commitments and some resourcing. Sceptics may be forgiven for considering the Dialogue another half-empty initiative unlikely to deliver such a result. But there is a silver-lining. Resolution A/RES/79/325 operationalises commitments to broadening access by establishing the Dialogue as a platform to address, among other things, “capacity gaps” and facilitate “access to artificial intelligence applications” as well as “high-performance computing and related skills in developing countries”. These are not aspirational generalities; they are agreed terms of reference for a recurring intergovernmental process. As Co-Chair Rein Tammsaar observed at the close of the inaugural session in Geneva, the Dialogue has already moved participants “beyond abstract debates towards discussions on [inter alia] public-sector capacity building”.
If the Dialogue is to justify its existence, it should function—at least—as a soft accountability mechanism: a forum in which commitments on technology access and capacity-building are not only announced but periodically measured against outcomes, with the second session in 2027 serving as the first meaningful checkpoint.
The Dialogue and (digital) human rights
Notwithstanding participants’ reluctance to invoke human rights in their statements, the Dialogue incorporates one thematic cluster dedicated to “respecting, protecting & promoting human rights”. This echoes the UN General Assembly’s focus on “respect for and protection and promotion of human rights in the field of artificial intelligence” when establishing the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance on 26 August 2025 (see Res 79/325 ¶ 4 (e)). Human rights also feature in the 2026 Preliminary Report of the Independent International Scientific Panel on AI (see pp 37-38), whose terms of reference were established by the General Assembly. This raises the question of whether the Dialogue is likely to advance the development of human rights principles and instruments specific to AI. Our initial assessment is that the rights-promoting impact of the Dialogue is likely to be limited. Treating human rights as merely one of the Dialogue’s seven thematic clusters allows for institutional compartmentalisation and permits avoidance by the unwilling or disengaged. This structure also enables Dialogue discussions on AI safety, innovation, interoperability and capacity-building to proceed predominantly in technical or economic terms, with human rights incorporated only retrospectively or marginally. So framed, human rights seem unlikely to inform every aspect of the Dialogue.
Some may argue that international human rights law (‘IHRL’) is not so easily dispensed with. Regardless of IHRL’s marginality in the Dialogue, it is nonetheless an established body of international law that binds States when they design, procure, regulate or deploy AI systems (see here and here). As such, its protections, including the rights to privacy, non-discrimination and equality, freedom of expression, due process, work, education and an effective remedy, apply throughout the AI lifecycle. For example, in considering how the right to non-discrimination applies in the AI context, the former UN Special Rapporteur on contemporary forms of racism, E. Tendayi Achiume, has observed that IHRL requires States to adopt a structural understanding of discrimination risks arising from AI and other emerging digital technologies, addressing not only explicit bias but also indirect and structural forms of discrimination (see A/HRC/44/57, 2020, ¶¶ 6, 38 & 48).
Through the UN Guiding Principles on Business and Human Rights (‘UNGPs’), IHRL also provides an authoritative framework for assessing whether companies developing and operating AI systems are meeting their responsibilities to respect human rights. These include responsibilities for undertaking human rights due diligence and providing or cooperating in remediation where harm occurs (see also A/79/170, 2024, ¶¶ 57(b)–(d) and 59(b)–(e)).
Even so, bringing IHRL to bear on AI development and deployment remains tremendously challenging. It arguably entails inventing an entirely new catalogue of rights – “an International AI Bill of Human Rights” – while also translating existing obligations into sustained institutional practices. Nothing about the Dialogue’s first session suggests that participants share an appetite to take up this challenge.
The elephants in the Dialogue room
With the Dialogue, the UN seems to want to nurture a broader set of shared understandings and common purposes in relation to AI than have emerged from the ad hoc international AI Summits convened by a succession of individual states since 2023 in Bletchley Park, Seoul, Paris, and New Delhi (with the latest planned for June 2027 in Geneva). Whether the UN can wrest momentum away from those summits will depend, however, on the combined influence of a herd of elephants milling about in the margins of the Dialogue. These concern the structure and operational practices of the International Telecommunication Union (ITU), which the Dialogue is convened alongside, and the position of the United States.
The relevance of the ITU hinges on it formally comprising the Dialogue’s joint secretariat. It also arises from the first of the Dialogues being held back-to-back with the ITU’s Artificial Intelligence for Good Global Summit. Similarly, the second Dialogue, in 2027, will be convened alongside the multi-stakeholder forum on science, technology and innovation for the Sustainable Development Goals (known as the STI Forum) which the ITU helps to plan.
The tenor and ambition of the Dialogue therefore bear the imprint of the ITU, the ‘most important UN agency you have never heard of’.
With reference to the potential governance implications of the Dialogue, two characteristics of the ITU warrant highlighting, in particular. First, unlike other UN agencies, the organisation has long admitted private corporations into its membership and decision-making processes. Corporate influence in the ITU was boosted by reforms of the mid-1990s that expanded the rights of private sector members. The second noteworthy feature of the ITU is the highly technical, specialised nature of much of its standard-setting work. This means that conflicts and tensions in and around the agency are mediated by mostly informal, expert-driven deliberations and rulemaking processes. It also means that the ITU’s ‘geopoliticisation’ proceeds without states’ uneven influence over ITU work becoming legible or contestable by the many public constituencies with a stake in it.
Even though the ITU was not on the list of international organisations from which the Trump administration vowed to withdraw the US in January 2026, long held US ambivalence towards the ITU nonetheless persists and extends to the Dialogue. In February 2026, White House technology adviser Michael Kratsios insisted that the US ‘totally reject[s] global governance of AI’. And although the US took part in the Dialogue, the remarks of White House Office of Science and Technology Policy Senior Adviser Katie Strickland championed national policymaking on the part of the Trump administration. This was in stark contrast to the remarks of UN Secretary-General António Guterres who cautioned that ‘[t]echnology that will shape the future of humanity must be shaped by all of humanity’, and not ‘by a handful of countries or companies’. More intriguingly, it is also at odds with the recent remarks of some major AI industry leaders including Microsoft Founder Bill Gates and Anthropic’s Darius Amodei.
Amid the dust that all these elephants kick up, it can be difficult to discern the extent to which the Dialogue will be crucial to these debates in future, or peripheral to them. Nonetheless, appetite for genuinely global dialogue, rather than a patchwork of regional initiatives, does appear to be rising.
Conclusion
The first Global Dialogue is best understood as an institutional opening, not a regulatory settlement. The Dialogue itself is not a negotiating forum and cannot adopt binding obligations. Its immediate output is a Co-Chairs’ summary rather than an agreed legal instrument. Those understandings will feed into the high-level review of the Global Digital Compact. Another limitation of the Dialogue was the lack of meaningful participation by marginalised communities. A civil society assessment of the inaugural Dialogue found that engagement with human rights issues remained uneven and that women, Indigenous Peoples and other affected communities were inadequately represented.
The question before international organisations, governments and civil societies is how to govern AI meaningfully, as Guterres remarked, a question that could not be more pressing. This Dialogue represents a modest yet important affirmation of the value of pursuing this question at a global scale, and the richness of the normative vocabulary available at that scale through which to pursue it.