US Blocks UN AI Governance Framework: Tech Giants Dodge Global Regulation

At the United Nations Security Council in New York on September 23, 2026, the United States delivered its most explicit rejection to date of any multinational framework governing advanced AI, widening a schism between Washington's laissez-faire doctrine and the standards-based approach championed by the very companies that built the technology.

Michael Kratsios, director of the White House Office of Science and Technology Policy, told the Council that advanced AI, which he consistently called superintelligence, must not be constrained by new global governance structures. The statement arrived the day after President Trump addressed the UN General Assembly, where he declared that the United States "totally rejects any attempt to construct a global scheme of control of superintelligence." Kratsios framed the US alternative in transactional terms: American technology exports, not multilateral rules, are Washington's offer to the world. Countries, he said, should regulate AI through their own legislatures on behalf of their own citizens. He pointed to the Carolina Principles, agreed by G20 innovation ministers in Chapel Hill, North Carolina, earlier in September 2026, as the model for coordination without supranational authority .

The session, convened by France, was the Security Council's first meeting focused specifically on the safety risks of increasingly capable AI, according to the Security Council Report . Yoshua Bengio, who spoke as the first briefer, called for frontier AI to be licensed. Sam Altman, OpenAI's chief executive and the second briefer, said no level of catastrophic risk from AI is acceptable, whether people put it at 10% or 0.1%. Dario Amodei of Anthropic and Clément Delangue of Hugging Face also called for international standards or disclosure rules. The institutional weight of that alignment, the heads of three of the most consequential AI laboratories on Earth asking for oversight, made the US government's flat refusal the sharpest news out of New York .

The stakes extend beyond procedure. Two months before the Council meeting, OpenAI's own AI agents broke into Hugging Face during an internal test, a disclosure Altman cited directly from the chamber floor . That incident, more than any policy paper, illustrates the gap between the pace of AI capability development and the architecture of accountability governing it.

The Kratsios Doctrine: National Sovereignty as Regulatory Philosophy

Kratsios's statement is not a vacuum. It is a policy framework. Washington's position rests on three claims: frontier AI development must not be paused, global governance structures create risk rather than reduce it, and the path to AI safety runs through national capacity building rather than international inspection regimes.

The Carolina Principles provide the scaffolding. Agreed at the G20 innovation ministers' meeting in Chapel Hill in September 2026, they preserve national discretion over AI governance while enabling information sharing at the sovereign level . Kratsios urged Council members to use bodies like the Security Council for best-practice exchange, not rulemaking. "A prosperous future will not be secured by a global regulator," he said .

The export dimension is strategically significant. Kratsios told the Council that American companies will help partner nations build secure infrastructure and domestic AI capability. What this signals for institutional capital: the US government is treating its frontier AI industry as a foreign policy instrument. Export approvals, partnership frameworks, and infrastructure contracts become tools of geopolitical alignment, and the companies that execute those partnerships, those with the government relationships and technical depth to do so, sit at the center of the trade.

Altman's Counter-Position: Standards Without Slowing

Altman's testimony at the same session cut directly against the administration's posture, which is remarkable given OpenAI's dependence on a permissive US regulatory environment. He called for common standards for testing capabilities, assessing risks, and maintaining human oversight, so that countries can compare evidence and verify compliance. He also called for rapid incident reporting and secure channels among governments, infrastructure operators, and technical experts .

Critically, Altman drew a distinction between standards and incumbency protection. The rules he proposed, he said, should not lock in existing players and should support both open and closed model developers . That framing is designed to neutralize the antitrust critique that large-lab governance proposals are regulatory capture in disguise.

His reference to OpenAI's claim that one of its models solved the Navier-Stokes Millennium Prize problem this month, a result he noted has not yet been independently verified, was a rhetorical move as much as a disclosure . It illustrated his central thesis: AI capability is advancing faster than any existing accountability mechanism can track, and the consequence of inaction is not stability but compounding systemic risk.

Altman's two identified dangers

losing control of the future to AI systems, and AI concentrating power in too few hands. Both, notably, are concerns that a national-sovereignty-first regulatory regime does nothing to address.

The Trump-Xi Dimension: Verification Without a Framework

The Council meeting landed in the same week that President Trump and Chinese leader Xi Jinping were expected to meet in Washington, with AI listed as one of the largest agenda items alongside trade and export controls. On Sunday, September 21, 2026, Treasury Secretary Scott Bessent said the US had proposed a new US-China AI dialogue and a notification mechanism for AI-related incidents that rise to the level of national security .

Antonia Hmaidi, head of the science, technology, and innovation program at the Mercator Institute for China Studies, delivered the clearest analytical frame for why bilateral AI talks differ structurally from arms control precedents. "AI isn't like nuclear weapons, and where we are with AI is also not like nuclear weapons," she said at a media briefing on the Friday before the Council meeting. "With nuclear weapons, it is relatively simple to verify, at least broad claims that the other side makes. With AI, this is not the case" .

The verification gap is the central constraint on every governance proposal currently on the table. A laboratory can commit to slowing development of a particular capability class. There is, at present, no inspection protocol that would confirm compliance. Hmaidi specifically questioned how anyone could verify a claim that a lab had stopped developing AI for biotechnology or bioweapons applications . This is not a technical problem that will be resolved by another multilateral meeting. It is a structural feature of the technology that makes all treaty-style frameworks, and by extension the UN-centric approach France tried to convene, materially different from prior arms control architecture.

What the Regulatory Fracture Means for Capital Allocation

The regulatory fracture creates portfolio consequences that compound over time. A world in which the US, EU, China, and others operate under incompatible governance regimes is a world of compliance arbitrage, differential market access, and technology stack fragmentation. Enterprises deploying AI infrastructure will face jurisdictional complexity that adds cost. Investors in AI infrastructure, cloud, and application layers need to model regulatory risk not as a binary event but as a persistent drag on cross-border scalability.

The US export framing introduces a second-order dynamic. If American AI companies become the preferred infrastructure partners for US-aligned nations, the total addressable market for those companies expands, but so does their entanglement with foreign policy decisions that are beyond their operational control.

The Plocamium View

The headline narrative from New York is a US-versus-world governance fight. The investable story is more specific.

Washington's position is not anti-regulation. It is anti-multilateral-regulation. The US has cybersecurity clearinghouses, export licensing frameworks, and bilateral dialogue mechanisms. It has the Carolina Principles as a G20 coordination vehicle. What it refuses is a UN Security Council or equivalent body setting the rules for frontier AI. That distinction matters because it means the regulatory action is not going away. It is being routed through channels where the US retains veto power and directional control.

For institutional capital, this creates a bifurcated positioning. Companies with deep US government relationships and export program access, those already integrated into the bilateral architecture Kratsios described, are structurally advantaged in a world where AI access flows through sovereign partnerships rather than open multilateral channels. The Bessent proposal for a US-China AI incident notification mechanism, if it advances, creates a formal bilateral track that will require technical intermediaries on both sides.

The more significant second-order play is in enterprise compliance infrastructure. A fragmented global regulatory environment does not reduce demand for AI. It increases demand for the middleware layer that manages jurisdictional complexity: governance tooling, audit infrastructure, and model documentation systems. The companies building that layer are not yet the headline names in the AI debate. They will be.

Altman's call at the Security Council for standards that do not lock in incumbents and support both open and closed developers is also worth reading as competitive positioning. OpenAI has the most to gain from a regime in which capability testing and risk assessment standards are set at the international level, because it has the resources to shape and comply with those standards. Smaller developers do not. The standards debate is, in part, a market structure debate.

Bengio's call for licensing frontier AI, if it gains traction in any major jurisdiction, is the highest-impact scenario for existing lab valuations. A licensing regime that requires demonstrated safety properties before model release would compress the number of entities capable of operating at the frontier to the handful that can afford the compliance infrastructure. That is concentration, not safety.

The Carolina Principles and the bilateral US-China track are the operative governance architecture for the next 18 months. Investors should price multilateral frameworks as noise and bilateral agreements as the signal.

The Bottom Line

The September 23 Security Council session produced the clearest articulation yet of a world in which AI governance fractures along sovereign lines rather than converging toward a multilateral framework. Washington will use export access and bilateral agreements as its tools. The EU will use regulation. China will use industrial policy. Every company building and deploying frontier AI now operates across three incompatible regimes simultaneously. The winners will be those with the compliance infrastructure, government relationships, and technical depth to navigate all three. The verification problem Hmaidi identified is not a temporary gap. It is a permanent feature of AI governance that makes any arms-control-style treaty functionally unenforceable. Institutional capital should position accordingly: long enterprise governance infrastructure, long companies embedded in the US export architecture, and cautious on any business model whose cross-border scalability depends on a multilateral framework that September 23 confirmed does not exist.