Altman and Amodei Take AI Governance to the UN Without Washington
The heads of OpenAI and Anthropic argued for cross-border coordination on AI risk at the UN, while the White House rejects international oversight — leaving US AI firms and their listed backers to navigate the…
OpenAI chief executive Sam Altman and Anthropic chief executive Dario Amodei used appearances at the United Nations to call for international cooperation on artificial intelligence, days after President Donald Trump dismissed international oversight of AI as a "globalist scheme" to control the technology.
The two men who run America's most closely watched artificial intelligence laboratories went to the United Nations to argue for something their own government has just publicly rejected: international cooperation on how AI is built and governed.
OpenAI chief executive Sam Altman and Anthropic chief executive Dario Amodei made the case for cross-border coordination on AI risk at the UN, according to CNBC Technology, days after President Donald Trump rebuffed what he characterised as a "globalist scheme" to control the technology. Altman is among several technology executives who have argued that AI companies should temper the pace of development in order to manage the risks that come with it.
That is an unusual configuration. The normal shape of a technology policy fight has industry resisting rules and governments imposing them. Here the largest private developers are asking for a governance framework at the multilateral level while the administration that hosts them is refusing to participate in one.
Two chief executives, one argument about speed
The substance of what Altman has argued publicly is straightforward, and it is the part of the debate that has the most commercial weight. If the pace of capability gains outruns the ability of the firms themselves — let alone regulators — to test and understand what their models do, then the risk is not a distant science-fiction scenario but a near-term operational one: systems deployed into finance, healthcare, defence and critical infrastructure whose failure modes are not fully characterised before they ship.
Amodei has built Anthropic's public identity around the same claim, positioning safety research not as a compliance cost but as the product. For both companies, an international framework has an obvious strategic appeal: a common floor of expectations applied to every serious developer, including those outside the United States, is more useful to a well-capitalised incumbent than a patchwork of national rules that vary by jurisdiction and leave competitors in permissive markets free to move faster.
Critics of that position note the same thing from the other direction. Rules written with the participation of the largest developers tend to be rules the largest developers can afford. The compliance burden of a safety-testing regime falls hardest on the newest entrants. Whether the UN pitch is public-spiritedness or moat-building is a question the frameworks themselves will eventually answer.
Why Washington's refusal matters more than the speeches
Multilateral AI governance without the United States is not governance. The overwhelming majority of frontier model training, the capital that funds it and the chip supply that enables it sits inside American corporate structures or under American export control. A UN process that the US declines to join can set norms, but it cannot bind the firms that matter.
The administration's framing — a globalist scheme to control the technology — is consistent with a broader position that has treated AI leadership as a strategic asset to be defended rather than a shared risk to be managed. It also fits a pattern in this White House's posture toward international bodies generally. For AI companies, the consequence is a widening gap between what their chief executives say in New York and what their home regulator is willing to sign.
That gap creates a practical problem. European and Asian regulators are not waiting. A company that commits to a voluntary international standard while its own government rejects the underlying framework ends up carrying two sets of obligations: whatever it has promised abroad, and whatever Washington decides domestically, with no guarantee the two align. Compliance costs rise; legal certainty does not.
What public-market investors are actually exposed to
Neither OpenAI nor Anthropic is listed, so there is no direct way to trade a view on this. The exposure runs through the listed companies that supply, fund and distribute their work — the chipmakers, the cloud platforms, the data-centre operators and the enterprise software firms building on top of frontier models. That chain is now a meaningful share of the market's index weight, which is precisely why a governance dispute between AI executives and the White House is a broad-market story rather than a sector one.
Neither OpenAI nor Anthropic is listed, so there is no direct way to trade a view on this.
The tape on Tuesday was soft across the board. At the last close, on Wed, 23 Sep 2026 at 20:00 GMT, the SPDR S&P 500 ETF Trust (NYSEARCA: SPY) finished at $767.81, down 0.72% on the day from a prior close of $773.38, with a day range of $766.50 to $773.05. The Invesco QQQ Trust (NASDAQ: QQQ), the tech-heavy Nasdaq 100 proxy, closed at $741.21, off 0.84% from $747.46 and the weakest of the three major benchmarks. The SPDR Dow Jones Industrial Average ETF (NYSEARCA: DIA) ended at $514.30, down 0.71% from $518.00.
Those are modest moves and nothing in them can be attributed to the UN appearances specifically. What they do show is the familiar ordering: when risk appetite softens, the Nasdaq proxy gives up slightly more than the broad index, because AI-linked names carry the heaviest weight there. Any regulatory outcome that changes the cost or the timing of frontier model deployment lands disproportionately on that same concentration.
The scenarios worth tracking from here
Three things determine whether this dispute stays rhetorical.
- Does the UN process produce anything binding? Declarations and working groups are cheap. A framework with reporting obligations, model evaluation requirements or incident disclosure is a different matter — and it is the one that would reach company income statements.
- Does Washington move from rejection to an alternative? Refusing international oversight does not preclude a domestic regime. The question for AI firms is whether the US writes its own rules or leaves a vacuum that states and courts fill unevenly.
- Do the labs act unilaterally? Altman's argument about tempering pace is, in principle, something a company can do without any regulator telling it to. Whether either firm slows a launch on safety grounds when a competitor does not is the test that will show how much of this is policy and how much is positioning.
For now the picture is a split screen: the people building the most capable AI systems asking the world to coordinate on limits, and the government with the most leverage over those systems declining to be part of the conversation. Markets are pricing the buildout. They have not yet had to price the rules.
Key facts
- Nasdaq 100 proxy (NASDAQ: QQQ): $741.21, -0.84%, last close Wed 23 Sep 2026 20:00 GMT
- S&P 500 proxy (NYSEARCA: SPY): $767.81, -0.72%, last close Wed 23 Sep 2026 20:00 GMT
- Who spoke: OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei, at the United Nations
- White House position: President Trump rebuffed international AI oversight as a 'globalist scheme' to control the technology
Frequently asked questions
What did the OpenAI and Anthropic CEOs argue at the UN?
Sam Altman of OpenAI and Dario Amodei of Anthropic pushed for international cooperation on artificial intelligence during appearances at the United Nations. Altman has been among several technology executives arguing that AI companies should temper the pace of AI development in order to manage the potential risks that rapid capability gains create.
What is President Trump's position on international AI governance?
Trump rebuffed international oversight of artificial intelligence, characterising it as a 'globalist scheme' to control the technology. That stance means the United States — home to the largest frontier AI developers and the chip supply chain behind them — is declining to participate in the multilateral framework the two chief executives were advocating at the UN.
Can investors buy shares in OpenAI or Anthropic?
No. Neither OpenAI nor Anthropic is a publicly listed company, so there is no direct way to take a position in either. Public-market exposure to frontier AI runs indirectly through listed chipmakers, cloud platforms, data-centre operators and enterprise software firms that supply, fund or build on top of their models.
How did US equity benchmarks close on the day of the report?
At the last close on 23 September 2026 at 20:00 GMT, the S&P 500 proxy SPY finished at $767.81, down 0.72%. The Nasdaq 100 proxy QQQ closed at $741.21, down 0.84%, the weakest of the three. The Dow proxy DIA ended at $514.30, down 0.71% on the day.
Why would AI companies want more regulation of themselves?
A common international standard applied to every serious developer can favour well-capitalised incumbents. It sets a predictable floor of expectations rather than a patchwork of differing national rules, and the cost of meeting safety-testing requirements weighs most heavily on newer, smaller entrants that lack the resources of established laboratories.
What should investors watch next in this dispute?
Three things: whether the UN process produces anything binding rather than declaratory; whether Washington offers a domestic alternative after rejecting international oversight; and whether the laboratories themselves slow a product launch on safety grounds when competitors do not. Each would change the cost and timing of frontier model deployment.
