20+ Nations Urge Human Control of AI as US Declares 'Super Intelligence' Race

For researchers, engineers, and product leaders, the safety-versus-speed debate has now been elevated to a head-of-state confrontation. A 20-plus-nation bloc is treating human control of AI as a non-negotiable design and deployment constraint, while the US is declaring 'super intelligence' an unbounded national race that rejects any globalist brake. The outcome will determine whether alignment work, open-weight releases, and frontier deployment rules diverge across markets — and which standards your models will ultimately have to meet.

At the United Nations General Assembly in September 2026, the governance of artificial intelligence hardened from a policy debate into an explicit geopolitical fault line between middle powers and global giants. On one side stood Australian Prime Minister Anthony Albanese, who joined more than 20 signatories — including Canadian Prime Minister Mark Carney, Spanish Prime Minister Pedro Sanchez, and German Chancellor Friedrich Merz — in a joint statement urging "extra safeguards" for AI and demanding it remain "firmly under human control." On the other stood U.S. President Donald Trump, who used his roughly 40-minute UN address less than 24 hours later to "totally reject any attempt to construct a globalist scheme to control the artificial intelligence being spoken of," propose renaming the technology "super intelligence," and declare, "Whoever wins super intelligence wins."

“

At the United Nations General Assembly in September 2026, the governance of artificial intelligence hardened from a policy debate into an explicit geopolitical fault line between middle powers and global giants.

The symmetry is deliberate and consequential. The middle-power statement, which explicitly "invited other nations to join," is an attempt to build a multilateral norm-setting coalition around human control of AI. It is led not by the EU institutions but by a cross-section of advanced democracies — including two NATO members (Canada and Germany, plus Spain) and two Five Eyes partners (Australia and Canada) — that have concluded they cannot win a pure compute-and-capital race against the United States or China, and therefore must compete on rules, safety, and legitimacy. For these countries, AI governance is a sovereignty question: the ability to shape how a general-purpose technology "rolls out," in Albanese's words, rather than having it "shape the way our society functions."

Trump's response reframes the debate entirely. By rejecting "globalist" control, renaming AI "super intelligence," and comparing it to the Industrial Revolution and the arrival of the internet, he positions the United States as a permissive, nationalist innovation zone where safety constraints are treated as both economically self-defeating and ideologically suspect. His rhetorical linkage of AI skeptics to climate-change warnings — "the very same people who said we'll all be dead in 12 years because of global warming" — signals that AI safety advocacy will be dismissed not as technical prudence but as political alarmism. The stakes, in his framing, are existential and zero-sum: whoever wins the race sets the terms for everyone else.

For the AI industry, the practical consequence is regulatory fragmentation at the highest level. A company building or deploying frontier models now faces a world where the United States signals minimal federal restraint and maximum acceleration, while a growing bloc of 20-plus nations organizes around safeguards, human control, and likely future procurement or market-access conditions. This is not merely a compliance nuisance; it can bifurcate release strategies, data governance, and safety investment. Companies may need to maintain different model versions, transparency disclosures, and deployment postures across markets, and the "invited other nations to join" language suggests the safeguards bloc will expand rather than contract.