The Quiet Compact

Most people I talk to about international affairs have not read the Global Digital Compact. This is not a criticism. It is a 36 page annex to a larger document, written in UN English, adopted at three in the afternoon on a Sunday in September 2024 alongside a much louder Pact for the Future. There was a small press release. A few think tanks weighed in. And then everyone moved on. I think that is a mistake. Because if you actually sit with the text, and read it the way you would read a contract or a constitution rather than a press release, the Global Digital Compact is the most consequential digital governance instrument since the WSIS outcome documents of 2003 and 2005. Almost no one has noticed yet. I want to do something simple in this piece. I want to walk through what the GDC actually binds, what it merely invites, and most importantly, what it carefully omits. Because the silences in this kind of document tell you more than the speeches. What the GDC binds The Compact sets out five objectives. Closing digital divides. Expanding inclusion in the digital economy. Protecting human rights online. Advancing responsible data governance. Enhancing international governance of AI. The honest answer is that most of these objectives are not binding in any enforceable sense. The Compact is a political declaration adopted by consensus, not a treaty. It commits signatories to "actions" rather than legal obligations. This is standard UN language and most readers know how to discount it. But there are two specific commitments inside the AI objective that are different from the rest. They commit member states to establish, within the UN system, two new mechanisms. An Independent International Scientific Panel on Artificial Intelligence. And a Global Dialogue on AI Governance. These are not aspirational. They are institutional. And as of 2026 they exist. The Panel was appointed in February 2026 with 40 members serving three year terms (UN General Assembly, A/RES/79/325). The Dialogue convenes for the first time in Geneva on 6 and 7 July 2026. So the binding part of the Global Digital Compact, the part that has actually happened, is the creation of two new UN bodies. Everything else is rhetoric until governments choose to act on it. This is important to understand before going further. The GDC did not regulate AI. It built the rooms in which AI may eventually be regulated. Those are different things and the distinction matters. What the GDC invites Most of the text falls into this category. The Compact "encourages" states to develop national AI strategies. It "supports" efforts to expand connectivity. It "welcomes" the contributions of civil society and the private sector. This language is the diplomatic equivalent of clearing your throat. It signals topics that future negotiations might address, without committing anyone to addressing them. But within this layer of soft commitments there are a few that deserve more attention than they have received. The Compact specifically commits to "support women and youth innovators and small and medium enterprises" in the digital economy. It commits to "strengthen legal and policy frameworks to protect children online." It calls for an "international scientific panel on AI" and a "global AI policy dialogue." The youth and children's references are worth noting because they map onto something concrete. The Italian youth delegate to the UN's Third Committee in October 2024 explicitly tied her national statement to the Global Digital Compact, citing the protection of "the rights of the child in the digital environment, with a specific focus on new AI implementations" (Italy, Third Committee Statement, 10 October 2024). That is the first instance I have found of a member state delegate using the GDC as a hook for AI advocacy. It will not be the last. The GDC has just provided a frame that any youth delegate, any minister, any civil society advocate can now use to anchor an AI argument inside an instrument that 193 countries adopted by consensus. That frame did not exist before September 2024. It exists now. What the GDC carefully omits This is the most interesting section, because it is where the diplomacy actually happens. The Compact does not contain the phrase "frontier AI." It does not mention model weights, training compute, or capability thresholds. It does not specify any AI system as posing systemic risk. It does not even use the phrase "general purpose AI," which is the operative term in the EU AI Act and most serious regulatory discussions. These are not accidents. These are negotiated absences. The countries that hold the largest stakes in frontier AI development, principally the United States and to a lesser extent the United Kingdom and a handful of others, did not want the GDC to constrain frontier development. The countries that worry most about frontier risk, principally several European states and a coalition of Global South voices, did not have the leverage to insist. The compromise was a document that talks about AI extensively while never mentioning the AI that most worries the people who actually understand the technology. This is not a failure of the GDC. It is what the GDC could achieve given the political constraints of 2024. But anyone reading the document for governance signals needs to see what is not there as clearly as what is. The Compact also says almost nothing about compute. There is no commitment on equitable access to AI infrastructure. No mention of the fact that frontier model training requires concentrations of compute that only a handful of states and companies can assemble. No acknowledgement that the geopolitics of AI is, in significant part, the geopolitics of GPUs. This silence is going to matter. Because while the Compact was being negotiated in New York, parallel developments were reshaping the actual landscape. The US-UAE AI Acceleration Partnership announced in May 2025. The Stargate UAE project, a one gigawatt cluster inside a five gigawatt campus in Abu Dhabi (G42 press release, 22 May 2025). The Saudi HUMAIN partnership with NVIDIA targeting 500 megawatts of capacity. The expansion of Chinese sovereign compute. The CHIPS Act money flowing through the US semiconductor ecosystem. None of this is captured in the GDC. The Compact talks about a digital world that is, by 2026, no longer the world we live in. The 2027 review The Compact specifies a high level review in 2027, during the 82nd session of the General Assembly (UN Pact for the Future, 2024). This is the moment when the omissions will either be addressed or they will calcify into the permanent shape of UN AI governance. What goes into that review is being negotiated now. The Geneva dialogue in July 2026 will produce a co-chair summary. The Scientific Panel will publish its first annual report in the margins of that meeting. Working groups will form between Geneva 2026 and New York 2027. And the states that show up to those working groups, that submit written inputs, that co-host side events, that build coalitions, are the states whose concerns will be reflected in the 2027 review text. This is where positioning matters. The middle powers I wrote about in an earlier piece are already positioning. The UAE submitted detailed written input to the Geneva dialogue consultation, which closed on 30 April 2026 (UN Global Dialogue, 2026 consultation portal). So did Estonia, El Salvador, Morocco, Singapore, the Netherlands, and a long list of others. Whether the United States, which voted against the Scientific Panel's establishment in February 2026, will engage substantively in the 2027 review remains genuinely uncertain. What this means if you are watching from outside If you are a young policy analyst, or a graduate student, or a civil society advocate, or a national official with no UN experience, here is what I would suggest looking at over the next eighteen months. First, read the Compact text itself. Not summaries. Not commentaries. The text. It is publicly available. It is shorter than most term papers. The 60 paragraphs on AI and digital governance can be read carefully in a single evening. Second, watch the Geneva dialogue's working groups. Their composition will tell you which states are serious about shaping the 2027 review and which are coasting. If the working groups end up chaired by middle powers, the review will look one way. If they end up dominated by superpower influence, it will look another way. Third, track the Scientific Panel's first annual report. Its tone, its scope, its courage in naming specific risks will set the template for everything that comes after. A report that hedges will signal a panel that has been captured by political concerns. A report that is genuinely scientific will signal an independent voice that may shape global discourse for years. Fourth, watch the youth track. The ECOSOC Youth Forum 2026 has explicitly included "Youth and AI" as a thematic area. National youth delegates will be carrying input papers into the Third Committee in October 2026. The youth statements made during those sessions are now, post-GDC, the most credible vehicle for raising frontier AI concerns inside a multilateral process. The Italian and Serbian precedents from 2024 are templates that others will follow. A closing note I started writing this piece because I was frustrated. Every public conversation about AI governance assumes that nothing has been agreed and nothing is being built. That is wrong. Something has been agreed. Something is being built. It is partial, it is flawed, it carefully avoids the hardest questions, and it is real. The Global Digital Compact is a quiet document. Its silences are louder than its commitments. Its real work will happen in the implementation, not the adoption. And the people who understand its text, who can read its omissions as clearly as its provisions, will have a quiet advantage in the conversations to come. That advantage is available to anyone willing to read. I would recommend reading. References United Nations. (2024, September 22). Pact for the Future, including the Global Digital Compact and the Declaration on Future Generations. A/RES/79/1. New York. United Nations General Assembly. (2025, August 26). Resolution A/RES/79/325 establishing the Global Dialogue on AI Governance and the Independent International Scientific Panel on AI. UN Secretary-General. (2026, February 12). Statement upon the appointment of the members of the Independent International Scientific Panel on AI. Retrieved from https://press.un.org/en/2026/sgsm23016.doc.htm United Nations. (2026). Global Dialogue on AI Governance: FAQ and consultation portal. Retrieved from https://www.un.org/global-dialogue-ai-governance/en/faq G42. (2025, May 22). Global Tech Alliance Launches Stargate UAE. Retrieved from https://www.g42.ai/resources/news/global-tech-alliance-launches-stargate-uae Permanent Mission of Italy to the United Nations. (2024, October 10). Statement to the Third Committee of the UN General Assembly, 79th Session. UN DESA Division for Inclusive Social Development. (2023, November 3). UN Youth Delegate Programme Update. DESA Voice. European Union AI Act, Regulation (EU) 2024/1689, Official Journal of the European Union, 13 June 2024. OECD. (2024). OECD AI Principles, updated version. OECD/LEGAL/0449.