NextFin News - The United Nations must open its doors to the artificial-intelligence industry or risk writing global AI rules that the people who actually build the technology will never follow, the President of the UN General Assembly said in an interview this week, turning private-sector engagement into the credibility test for the world body's bid to govern the defining technology of the decade.
Khalilur Rahman, the Bangladeshi diplomat who took office as President of the General Assembly on 8 September, framed the challenge as a choice between relevance and ritual. The UN has spent the past two years assembling an AI governance architecture - an Independent International Scientific Panel on AI, a Global Dialogue on AI Governance, and the 2024 Global Digital Compact - but the capital and the engineers that determine how those rules land in practice sit largely outside the tent.
The timing is deliberate. Rahman's first major test arrives with High-Level Week in New York, 21-25 September, when heads of state and government converge on Midtown Manhattan for the annual General Debate that opens on 22 September. Artificial intelligence is one of the six priorities he has set for the 81st session, listed explicitly as "governance of emerging technologies, including artificial intelligence." His message to the industry is that the UN cannot be a serious player in AI governance unless the industry shows up as a partner rather than a spectator.
The Machinery Exists. The Builders Are Mostly Absent.
The question the UN faces is not whether it has institutions. It does. On 26 August 2025, the General Assembly adopted Resolution A/RES/79/325 by consensus, formally establishing two new bodies: the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance. The Panel - the first global scientific body dedicated entirely to artificial intelligence - brings together 40 experts serving in their individual capacity, drawn from all five UN regional groups. The General Assembly voted to appoint the 40 members on 12 February 2026, with 117 member states in favour; the United States voted against, joined by Paraguay, while Ukraine abstained.
The Panel held its inaugural meeting on 3 March 2026 at UN Headquarters in New York and delivered its preliminary report in time for the Dialogue's first session, held 6-7 July 2026 in Geneva. Outgoing Assembly President Annalena Baerbock, addressing the opening, called the panel's creation a milestone.
The establishment of the independent international scientific panel on artificial intelligence is therefore a major milestone for the first time delivering parity across genders and regions while drawing from expertise from all sectors.
Nor is the ambition in doubt. The Dialogue was conceived as a universal forum - the UN's education agency describes it as the platform "where all governments, private sector, academia and civil society will convene" - with a second session already scheduled for May 2027 in New York. Yet the design and the participation are not matching. The two actors whose behavior determines whether any global framework bites - the United States and the European Union, and the companies spending hundreds of billions on AI infrastructure - have kept their distance. Richard Gowan, an analyst tracking the process, put it plainly ahead of the New York gathering.
We know that the US and EU members are not really engaging with this very much. They don't want UN regulation of AI, and instead we're seeing Anthropic ... OpenAI ... a bunch of companies in the private sector doing their own events about the future of artificial intelligence.
This is the gap Rahman is trying to close. A rulebook that the rule-makers ignore is not a rulebook; it is a statement of aspirations. The Assembly President's argument - that the UN must unlock its doors to the AI industry - is an admission that the organization cannot regulate from the outside what it cannot convene from the inside.
The Real Fight Is Over Who Governs the Capital
Beneath the diplomatic language lies a harder question: who governs the money? The Independent International Scientific Panel on AI's preliminary report, published in July 2026, quantified the concentration at the heart of the problem. Hyperscaler capital expenditure on AI has risen about five times since 2023, from roughly $150 billion to a projected $770 billion in 2026 - around three times the combined spend across the rest of the world. The United States hosts close to 75% of global AI compute capacity. Meanwhile, leading AI companies' annualized revenues have risen more than thirtyfold over the same period, from about $2 billion in 2023 to more than $70 billion in 2026.
That concentration is not just an economic statistic; it is a governance fact. When three times as much AI investment flows through the United States as through every other country combined, the American regulatory perimeter and the self-governance choices of a handful of labs already function as the de facto global regime - regardless of what any Geneva forum resolves. Standards-setting is a network-effect game: the framework that attracts the most capital and the most engineering talent becomes the operating system everyone else must interoperate with. A UN process that the capital ignores does not fail loudly; it simply becomes irrelevant by default.
The UN's counter-offer is legitimacy and inclusion. Secretary-General António Guterres has argued that global checks and balances on AI should include "locked-in access to the self-learning tech for developing countries," while all AI data centres should be powered by renewable energy by 2030. In a report to the General Assembly released 11 September 2026, he recommended that member states advance the establishment of a Global Fund for AI Capacity-Building, building on his 2025 proposal for innovative voluntary financing options to address compute, data, skills, and governance gaps in developing countries. At the AI Impact Summit in New Delhi in February, he put a figure on the ask: a US$3 billion global AI fund aimed at equipping developing nations with skills, data access, affordable computing, and inclusive frameworks.
Guterres framed the choice in stark terms at the Dialogue's opening in Geneva.
Used well, and shared widely, AI could compress decades of development into years. It could become the great equalizer of the 21st century. But no future builds itself. And so the choice before us is not between faith in AI or fear of it. It is between governing by design and drifting by default.
The second-order implication is what makes Rahman's intervention significant for markets, not just diplomats. If the UN succeeds in pulling the industry inside, the likely outcome is not a single global regulator - that is politically impossible - but a layer of interoperability: common disclosure standards, shared safety benchmarks, and compatible rules that reduce the cost of complying with multiple regimes. That is valuable to any company operating across borders. Guterres made the economic logic explicit in Geneva.
When countries align on how to test systems, measure risk and assign responsibility, safety travels with the technology. When they do not, a patchwork of incompatible rules raises costs, divides the world - and protects no one.
If the UN fails, the industry gets what it appears to prefer: fragmented national rules, negotiated one capital at a time, with the largest players writing much of the text themselves.
This Is a Regime Shift, Not a Cycle
It is tempting to read the UN's AI scramble as cyclical - a belated institutional reaction that will normalize once the technology matures. That reading is wrong. Three structural features make this a regime shift rather than a cycle that will revert on its own.
First, the concentration of compute is self-reinforcing. A $770 billion annual capex base, three times the rest of the world's combined, does not disperse through market forces; it compounds through economies of scale, talent agglomeration, and the increasing capital intensity of frontier models. Second, the geopolitical fragmentation of technology standards is a political choice, not a market fluctuation - major powers are actively building separate rulebooks, and the US vote against the scientific panel's membership is evidence that even the UN's most modest coordination mechanisms face organized resistance. Third, the UN itself has now institutionalized the response: two new standing bodies, a compact adopted by member states in 2024 as part of the Pact for the Future, and a Secretary-General whose term ends with AI governance as one of his final priorities.
The practical consequence is that the question is no longer "will the UN catch up." It is which of the competing governance layers - the UN's inclusivity-and-legitimacy layer, the American regulatory perimeter, the European rulebook, or corporate self-governance - becomes the binding one for the capital that actually matters. That is a structural contest, and it will be decided by behavior, not by resolutions.
The Counter-Thesis: The UN Never Meant to Be a Regulator
The strongest argument against Rahman's framing is that it misstates the assignment. The Global Dialogue was never designed to be a regulator. The guidance prepared for the Dialogue describes its purpose as "stitching together fragmented national, regional, and private-sector efforts into a coherent governance architecture" - an interoperability exercise with "formal space for states, expert communities, civil society and the private sector to contribute." On this reading, private-sector participation is already baked into the design; companies are meant to be stakeholders, not subjects. The Dialogue's job is coordination, not control.
There is force in that objection. A UN forum that tried to license frontier models would be dead on arrival in the capitals that matter. The organization's comparative advantage is exactly what Rahman is emphasizing: convening power, legitimacy for developing countries, and a neutral table for technical interoperability. The guidance document itself warns that technology companies "need a seat at the table as active stakeholders, not bystanders offering voluntary disclosure" - which is precisely the role the architecture anticipated.
But a sound design without participants is decoration. The counter-thesis holds only if the private sector actually uses the space it has been given. So far, the evidence points to bystanders. If the labs and the regulators continue to hold their own events on the margins of the General Assembly rather than inside the Dialogue, then the "unlock the doors" plea is not a refinement of a working design - it is a diagnosis that the design is failing in its first year. The US vote against the panel, and the absence of US and EU engagement that analysts have noted, are the behavioral evidence - and behavior trumps design documents.
What to Watch: The Signals That Will Settle the Argument
The base case is that the Dialogue becomes a real but limited interoperability forum. Major labs engage selectively - enough to claim participation, not enough to submit to binding oversight - and global rules remain fragmented but somewhat more compatible. Developing countries gain a voice they lacked, but the binding decisions continue to be made in Washington, Brussels, and corporate boardrooms.
The upside case requires a behavioral shift: the United States and the European Union re-engage with the UN process, the major AI companies commit to the Secretary-General's transparency initiative and the 2030 renewable-data-centre disclosure, and the proposed Global Fund for AI attracts real capital. That would produce a genuinely inclusive regime with teeth in the form of disclosure rather than licensing.
The downside case is a return to the pattern described ahead of this week's gathering: the UN process becomes a talking shop for the Global South while the real rule-making happens elsewhere. In that world, Rahman's warning becomes a footnote to a year in which the industry consolidated its freedom of action.
The falsifying signal is concrete and observable. If the Dialogue's next session - scheduled for May 2027 in New York - passes without participation from the major AI labs, without renewed US and EU regulatory engagement, and without any major company committing to the 2030 renewable-data-centre disclosure, then the premise that the industry wants a seat at the UN table is wrong - and the "unlock the doors" argument collapses into a plea that no one answered.
The calendar adds urgency. Guterres's term as Secretary-General ends on 31 December 2026, and Rahman's presidency will oversee the selection of his successor - a process that will shape the UN's institutional direction for years. The AI file is one of the outgoing chief's signature priorities; whether it survives the transition is itself a test of whether the UN has moved AI governance from aspiration to institution.
Rahman set the tone for his presidency on 8 September with a line that applies directly to this file.
Trust cannot simply be requested. It must be earned. It will be earned through decisions we take, the commitments we honour and above all, the results we deliver for the people we serve.
The AI industry will be watching whether the UN can deliver something it values enough to walk through the door for.
The UN can write the world's most inclusive AI rulebook. It just needs the people who own the compute to show up - and the companies that control the capital have so far preferred rooms where they hold the keys.
Explore more exclusive insights at nextfin.ai.
