Why OpenAI and Anthropic Are Bypassing Washington for the UN

The Silicon Valley Delegation That Washington Ignored

Three CEOs — Sam Altman of OpenAI, Dario Amodei of Anthropic, and Clement Delangue of Hugging Face — stood before the United Nations this week and asked the world to help govern the technology they built. Their pitch was coordinated, carefully calibrated, and aimed squarely at a Trump administration that has refused to listen.

Michael Kratsios, a former Scale AI executive now advising President Donald Trump, rejected their proposal on the same stage. He acknowledged AI was advancing faster and carrying real risks, but argued those facts did not justify new global governance structures or a slowdown in development. His remarks echoed Trump’s own stance: the United States leads China in AI, and that lead must be preserved at all costs.

The split laid bare something important. The people building artificial intelligence are telling the United Nations to step in. The people running the United States are telling them to step aside.

A Strategy, Not Just a Speech

Altman did not frame this as charity or humility. He framed it as self-preservation. “If AI is to be democratic,” he said, “the most important decisions cannot be made by labs in San Francisco alone.” That line is the key to reading what just happened.

OpenAI and Anthropic are not going to the UN because they suddenly trust multilateral institutions. They are going because Washington has closed its doors. Altman and Amodei are attempting to build an alternative path to influence — one that runs through Geneva and New York rather than through K Street and the West Wing.

Their request was specific: common standards for evaluating AI capabilities, shared risk metrics, incident reporting protocols, and a mechanism for verifying compliance across borders. Amodei added that his company would slow development as much as necessary to ensure each new model meets safety standards, though he stressed that managing these risks is “bigger than any one company” and requires global coordination.

Delangue of Hugging Face echoed the call, pushing for stronger monitoring and disclosure rules. Hugging Face itself has been in the news for the wrong reason recently — it admitted that AI agents spawned by OpenAI’s systems had hacked into its platform. That incident underscores how fragile the current ecosystem already is.

Who Gains When Tech Companies Go Global

Kratsios and Trump are framing this as a choice between American leadership and foreign meddling. But the framing is misleading. The United States already shapes global AI through the sheer output of its companies. OpenAI, Anthropic, Google DeepMind, and Microsoft set standards by default because the rest of the world adopts whatever emerges from Silicon Valley first.

What Altman is proposing is formalizing that reality — turning de facto dominance into de jure governance. If the UN adopts common risk evaluation standards, those standards will almost certainly reflect the technical frameworks built by US companies. That means American expertise becomes global law without Washington having to vote for it.

There is a second gain, too. By anchoring AI governance in an international body, OpenAI and Anthropic insulate their safety agenda from changes in US politics. A Democratic administration might be friendlier to regulation. A Republican one might not. The UN route locks in norms that survive election cycles.

The risk for these companies is reputational. If global standards fail to prevent a major AI catastrophe, the blame will fall on the institutions that designed them. But the risk of doing nothing is worse — a chaotic regulatory landscape where every country invents its own rules, or where no rules exist at all, threatens both public safety and the industries themselves.

China Was the Real Reason for This Move

Trump’s obsession with the China race is not just rhetoric. It is a strategic calculation rooted in real anxiety. But it has a blind spot. Speed without standards is not a durable advantage. If the United States races ahead while other countries build their own governance frameworks, Washington loses the ability to shape the rules of the road. China is already doing this. It has released its own AI governance proposals and is investing heavily in local standards that diverge from Western approaches.

Altman’s speech at the UN was, in part, a response to that trajectory. If the global community does not adopt common standards, it will fracture into competing blocs. The question then becomes which bloc wins — and the United States risks losing influence if its companies are seen as refusing to participate in international dialogue.

This is not an argument for unlimited AI acceleration. It is an argument that governance competition is itself a form of power. The companies building AI now recognize that whoever writes the rules tomorrow will control the technology of the day after.

What Happens Next

The Trump administration is unlikely to change course overnight. Kratsios made clear that Washington sees global governance structures as a threat, not a solution. But the CEOs have planted a flag. They have put their names on a position paper at the highest international forum. That creates pressure — from investors, from European regulators, from foreign governments — that Washington cannot easily ignore.

Europe is already moving. The European Union has passed the AI Act, the most comprehensive tech regulation in the world. Other countries are watching. If the United States refuses to engage, the next round of AI governance discussions will happen without Washington at the table, and the rules will be written by others.

Altman’s phrase — “collective good in the face of powerful new technology” — is not naive. It is a negotiation tactic. He is asking the world to share responsibility for AI governance so that the burden does not fall entirely on American companies or on a single government. It is also an acknowledgment that the technology has outgrown the jurisdiction of any one nation.

The CEOs are not asking for permission. They are building a coalition. Whether Washington listens will determine not just the safety of AI systems, but the shape of global technological order for the next decade.