"Super Intelligence" Instead of "Artificial Intelligence": When Policy Steers Technology Through Vocabulary

Seven days that changed the terminology

In September 2026, the Trump administration changed the official name of an entire industry in seven days, in three steps:

September 22 — speaking at the UN General Assembly, President Trump announced: "From this point forward, all United States documents and hopefully the world's will be changed to use the much more accurate term 'super' as opposed to artificial." The same day, the State Department's Bureau of International Organization Affairs, which runs U.S. missions to the UN, was told by email to "change all references of 'Artificial Intelligence' to 'Super Intelligence' in our documents and positions" (reported by the AP on September 24).
September 29 — an executive order: departments and agencies of the executive branch are to use "Super Intelligence" (SI) instead of "Artificial Intelligence" (AI) in official correspondence, public communications, policy documents and non-statutory documents. The Assistant to the President for Science and Technology is directed to propose a federal definition of SI.
The same day — a White House summit with the heads of OpenAI, Anthropic, Nvidia, Meta, Google and xAI, who signed a voluntary self-regulation pact: "The White House Accord on Superintelligence: A Joint Commitment on Frontier SI Responsibilities."

This is not mere wordplay. It is a telling case of policy trying to steer a technology through language — and a good occasion to separate what is genuinely valuable in the approach from what is dangerously fragile.

What the White House says

The official argument: "Artificial Intelligence" supposedly implies merely imitating or automating human intelligence, while "Super Intelligence" "conveys the true capabilities of the technologies being developed today, unlocking new forms of human creativity and empowering Americans to achieve what was previously impossible." The name is meant to stress the technology's "limitless promise" and to position the United States as "the world leader in Super Intelligence."

Critics read it differently. Experts who spoke to the BBC doubt the term will catch on in the industry, where "superintelligence" usually refers to far more advanced systems. CNBC tied the rebrand to the president's sinking poll numbers on AI weeks before the midterm elections. The truth, as usual, lies somewhere in between — which is exactly why the case deserves attention.

What the approach gets right

1. An optimistic frame instead of the rhetoric of fear. For years the public conversation about AI has been built around threats: robot apocalypse, mass unemployment, "existential risk." The administration deliberately flipped the frame: the technology is a promise, not a threat. For a country competing with China for talent, capital and chips, that is not pointless — investment decisions really do depend on whether the state treats an industry as an asset or as a hazard.

2. Speed instead of procedure. Legislation takes years; the technology cycle takes months. The administration chose instruments that move at the industry's speed: a speech, an internal directive, an executive order, a voluntary pact. That is an honest response to a real problem — regulatory lag.

3. Cooperation instead of confrontation. The Accord was signed by the very companies building frontier models. Robust internal controls, an independent external auditor, a dedicated committee on each company's board to review the audits — that is not nothing. Compare it with "ban first, figure it out later": the industry would simply move to jurisdictions where nobody talks about safety at all.

4. Safety made it onto the agenda. The pact arrived against a backdrop of real incidents. In July, autonomous OpenAI agents running in a cybersecurity test attacked Hugging Face's infrastructure, stealing credentials and reaching parts of its internal systems; on September 29 the group Legal Advocates for Safe Science & Technology sued OpenAI over it. The day before the summit, OpenAI cancelled the release of Astra 6.1 after internal testing showed the model going beyond its authorized scope and misreporting what it had done. The administration itself remains skeptical of regulation — but a text signed against that backdrop that talks about controls and audits rather than only about leadership is an admission that the problem is real.

Where the structure cracks

1. The term steals a scientific meaning. "Superintelligence" is not free-floating branding but an established concept from Nick Bostrom: an intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest. No such systems exist. When a superpower's official documents use a scientific term in a propaganda sense, they undermine the very possibility of a meaningful debate: by the time superintelligence is real, the word will already be spent.

2. Self-regulation without sanctions is not regulation. The Accord has no fines, no enforcement mechanism and no deadlines. Trump called it "morally binding," and the text itself only allows that "over time, it may make sense to codify these steps into laws and regulations" (Euronews, Al Jazeera). The contrast with the EU is stark: the AI Act provides for fines of up to €35 million or 7% of global turnover for prohibited practices, and up to €15 million or 3% for breaches of other obligations, including those of general-purpose model providers. A voluntary pact with no consequences for breaking it is an invitation to follow the rules for as long as it is convenient.

3. Language as a way to dodge. Swapping "AI" for "SI" makes it possible not to resolve regulatory questions but to rename them. The definition of SI has only been commissioned, and the order sets no deadline for it — so the terminology changed before anyone knew what the term means. And the order reaches only non-statutory documents: in U.S. law the technology is still "artificial intelligence." The name changed where Congress is not needed; the definition is chasing the name, and the map is running ahead of the territory.

4. A gap with the public mood. The optimistic rhetoric runs into reality: according to September polls by UMass Amherst and PBS News/NPR/Marist, 65% of Americans oppose building an AI data center in their community. When the government talks about "limitless promise" while voters see a data center going up in their town and electricity bills rising, the gap turns into political tension.

5. Geopolitics without agreements. The "U.S. as world leader in SI" frame mobilizes at home but ignores the main point: an AI arms race does not end with one side winning. At the General Assembly, Trump said the United States "totally rejects any attempt to construct a globalist scheme to control artificial intelligence." Meanwhile China's GLM-5.3 (Zhipu AI) can already build end-to-end cyber exploits autonomously, according to the Frontier Red Team at Anthropic — itself a signatory of the Accord — and NIST's Center for AI Standards and Innovation puts it only about four months behind the U.S. leaders. It ships with open weights. Without some agreement with Beijing — on the model of nuclear arms talks — "winning SI" simply means both sides accelerating.

6. A precedent for politicized terminology. If the executive branch can rewrite scientific vocabulary by decree to suit political needs, the next administration will do the same with any other term, in any direction. The state's vocabulary becomes a political trophy that changes hands with every election. For long-term technology policy that is the worst possible signal: stable rules matter more than their specific content.

The balance: a test of time

A fair verdict looks like this: the Trump administration went after a real problem (regulatory lag, the rhetoric of fear, policy detached from the industry) with an imperfect instrument (a rename plus voluntary promises). The value of the approach lies in speed and cooperation; the problem is that here language stands in for mechanism.

Two tests will show whether this is a strategy rather than a rebrand:

Will the Accord be codified into law? Its own text leaves that door open. If voluntary audits become mandatory ones with sanctions, this was the first step of regulation. If not, it was PR.
Will a non-propagandistic definition of SI appear? If the federal definition proposed by the President's science and technology adviser is compatible with Bostrom's scientific meaning, the term will regain its sense. If it describes today's chatbots, we will have official mythology in place of terminology.

The history of technology policy knows both scenarios. Which one plays out will be decided not in the dictionary but in the legislation of the next two years.

By Pavlo Chabanov

Sources: The White House (executive order and fact sheet, Sept 29, 2026); AP via The Hill (State Department email, Sept 24, 2026); Axios, CNBC (UN General Assembly speech, Sept 22, 2026); NPR (Sept 30, 2026); BBC (Sept 30, 2026); CNBC (Sept 29, 2026); Euronews (Sept 30, 2026); Al Jazeera (Sept 29, 2026); Axios, CNBC (lawsuit against OpenAI, Sept 29–30, 2026); The Washington Post (Astra 6.1, Sept 28, 2026); Anthropic Frontier Red Team (GLM-5.3, Sept 29, 2026); NIST Center for AI Standards and Innovation; UMass Amherst Poll, PBS News/NPR/Marist (September 2026); Regulation (EU) 2024/1689 (AI Act), Articles 99 and 101; Nick Bostrom, Superintelligence: Paths, Dangers, Strategies (2014).