Trump Wants an AI Army. The UN Wants AI to Stay Human-Controlled.

The UN Is Trying to Put the Brakes on AI. The White House Wants to Floor It.

The 81st session of the UN General Assembly opens this week in New York, and one of the defining questions of the decade is sitting on the agenda: should humanity pace the development of artificial intelligence, or race ahead?

The answer depends entirely on where you stand. In the chamber, the narrative coming out of the UN Secretariat, France’s Security Council presidency, and Big Tech’s safest voices is clear — AI could slip beyond human control, and the world needs guardrails before it’s too late. In Washington, the signal is the opposite.

Who Is Speaking, and What Are They Saying?

UN Secretary-General António Guterres told reporters on September 16 that the warnings coming from the front lines of AI development cannot be ignored, and that AI safety will be a central topic among world leaders next week. That is no small framing. Guterres has positioned himself as a skeptic of unfettered technological acceleration, and his language — calling for information-sharing among the countries best positioned to develop AI, for transparency mechanisms, for a multilateral system that prevents risk — signals that the UN is taking this seriously.

France, holding the Security Council presidency for September, has convened a meeting for September 23 specifically to analyze what it calls the risk of losing control over AI and the danger that advanced AI models could be used to develop, promote, or carry out actions threatening international peace and security. That is direct language for a body that rarely uses such framing.

Sam Altman, CEO of OpenAI, is set to brief that Security Council meeting. So is Dario Amodei of Anthropic. These are not regulators or diplomats. They are the people building the systems they are now warning about. Their presence at the UN is itself a story — it signals that the industry’s cautionary voice has crossed from press releases into the highest corridors of international security deliberation.

On September 24, a side event organized by AI Safety Connect — a coalition co-founded by governance expert Nicola Miayu — will bring together the UN Development Programme, UNESCO, and member-state representatives to discuss pre-release safety testing, independent third-party evaluation, global common safety standards, and international coordination protocols for AI crises.

The architecture of a safety regime is being sketched in real time. The question is whether it will hold.

The Trump Contradiction

Here is what the UN coverage largely omits: Donald Trump has spent his second term pushing a radically different vision. He has advocated for an AI-powered military, spoken of creating an AI czar within the executive branch, and framed American leadership in artificial intelligence as a matter of national survival against China.

This is not a side note. It is the central contradiction of the moment.

The UN debate treats AI safety as a collective action problem — the kind of challenge that requires shared standards, transparency, and multilateral oversight. Trump’s AI-military vision treats it as a zero-sum competition — the kind of challenge where speed matters more than guardrails, and where ceding any ground to rivals is an existential risk.

These two frameworks cannot coexist without friction. If the United States is simultaneously advocating for AI safety standards at the UN and pouring resources into an AI-driven defense posture, the rest of the world will read that as hypocrisy. China will read it as proof that America wants rules only when they constrain others. Europe will read it as evidence that Washington cannot be trusted with global AI governance leadership.

Why This Collision Matters Beyond the UN Halls

The stakes are not abstract. The source material notes that Trump and Chinese President Xi Jinping are expected to hold bilateral talks following the UNGA high-level week. AI will be on the agenda. Whatever emerges from that conversation will set the tone for whether the safety frameworks being discussed in New York have any chance of translating into enforcement.

If the United States and China cannot find common ground on AI standards — and most analysts expect them not to — then the UN’s safety architecture risks becoming a voluntary exercise with no enforcement mechanism. That is exactly what Miayu warned about: the harder part is not agreeing on what the framework should look like, but actually implementing it.

China’s position is clear from the source: it accuses the United States of seeking to monopolize AI technological leadership. The United States’ position, as reflected in its military AI push, validates that accusation in Beijing’s eyes. The cycle feeds itself.

Who Wins and Who Loses

If the multilateral safety framework gains traction, the winners are smaller nations that lack AI development capacity but bear disproportionate risk from autonomous weapons, deepfake-enabled disinformation, and unchecked algorithmic decision-making. The losers are those who benefit from unregulated acceleration — primarily the well-capitalized actors on both sides of the Pacific who see guardrails as competitive disadvantages.

If the framework falters, as seems increasingly likely given the US-China impasse, the world moves toward a de facto AI arms race with no diplomatic circuit breakers. Corporate safety pledges become marketing. The UN becomes a talking shop. And the people who built the most powerful systems — Altman, Amodei, and their counterparts in China and elsewhere — answer to no one.

What Happens Next

Watch two things. First, whether the Security Council meeting produces any joint statement or commitment beyond rhetoric. Blank resolutions on AI safety have historical precedent — they look like progress and mean almost nothing. Second, watch the Trump-Xi talks. If they produce even a modest agreement on information-sharing or crisis communication channels around AI, it would be the most significant outcome of the entire UNGA session. If they produce nothing, the safety architecture discussed in New York will remain exactly what it is today: a proposal without a enforcer.

The UN is trying to solve a coordination problem with voluntary cooperation. The White House is operating in a competition framework. Until those two realities converge — or one defeats the other — AI governance will remain a debate, not a system.