Trump Wants to Rename AI “Super Intelligence.” The Human-Control Question Matters More.
Share
Donald Trump’s September 22, 2026 UN speech turned an already loaded technical term into a political branding proposal. The important question is not whether the name sounds better. It is whether calling today’s AI “super intelligence” blurs distinctions that matter for safety, regulation, accountability and human agency.
The short answer
Trump did say the United States would start referring to artificial intelligence as “super intelligence” in government documents. But in AI research, “superintelligence” already has a much narrower meaning: systems that substantially exceed human-level capability, not a new marketing label for the AI tools people use today.
That distinction matters because some of the leading AI labs and researchers using the term to describe future systems beyond AGI have also warned that such systems could require unusually strong safeguards. Meanwhile, the Trump administration’s own June 2026 national-security AI directive says advanced systems should remain steerable, controllable and accountable.
My concern is not that the rename suddenly makes the technology more powerful. It is that language can change how people understand who has power, who has responsibility and how much human direction still matters.
What did Donald Trump actually say about “super intelligence”?
On September 22, 2026, during his address to the United Nations General Assembly, Trump said the United States would use the term “super intelligence” instead of artificial intelligence in government documents. In the same section of the speech, he rejected calls for broad new restrictions on AI, argued that the technology could be larger than the Industrial Revolution and said the Justice Department could step in if companies crossed lines that required intervention.
That combination matters. This was not simply a naming joke. It appeared inside a larger policy argument about growth, regulation, national competition and AI risk.
Read Reuters’ report on the September 22 speech →
There is also an important verification point. At the time of publication, I have not found a new executive order or federal rule that legally redefines the entire category. Existing White House policy documents still use the term artificial intelligence. That means the speech is a real presidential statement, but formal implementation across agencies still needs to be watched.
The problem: “superintelligence” already means something
This is the part of the story that should not get lost in the politics.
In ordinary AI policy, artificial intelligence is a broad category. NIST includes definitions describing AI as machine-based systems that operate toward human-defined objectives and produce predictions, recommendations or decisions. That definition covers a huge range of systems, from narrow machine-learning tools to modern generative models.
See NIST’s artificial-intelligence glossary →
Superintelligence is different.
OpenAI has described superintelligence as future systems dramatically more capable than AGI. In its 2023 governance paper, OpenAI explicitly argued that superintelligence should not be treated like lower-capability technology and would require special coordination and oversight.
Read OpenAI’s “Governance of superintelligence” →
Google DeepMind made the distinction even more current in June 2026. Its paper From AGI to ASI describes artificial general superintelligence as a system more intelligent and cognitively capable than large organizations of humans.
Read Google DeepMind’s 2026 AGI-to-ASI paper →
Why the terminology matters
If “AI” is the umbrella category and “superintelligence” is a future capability threshold, using the second term as the everyday name for the first collapses an important distinction. A chatbot, an image generator, an AI music tool and a hypothetical system exceeding large groups of expert humans would all be linguistically flattened into the same label.
Trump’s own AI policy makes the rename more interesting
This is where the story becomes more complicated than “Trump is anti-regulation.”
The administration has consistently favored faster AI deployment, fewer barriers and a national policy designed to protect U.S. technological leadership. The 2025 AI Action Plan explicitly prioritized accelerating innovation and reducing what it described as burdensome regulation.
Read America’s AI Action Plan →
But the administration has also used very specific human-control language. On June 5, 2026, Trump signed a national-security AI memorandum directing that deployed systems be robust, steerable, controllable and preserve clear lines of accountability.
Read the White House national-security AI fact sheet →
That language is closer to the principle I care about: powerful AI should not weaken human responsibility. It should increase the need for clear responsibility.
So the tension is not simply between Trump and AI safety advocates. There is a tension inside the administration’s own framing: the policy says control and accountability matter, while the new label rhetorically elevates the machine itself.
Why the current AI-risk debate makes this timing important
The term “superintelligence” is not neutral because AI labs themselves have spent years connecting it to extreme-risk scenarios.
OpenAI’s former Superalignment project said humans would need technical breakthroughs to reliably steer systems much smarter than us. Anthropic continues to maintain a Responsible Scaling Policy designed around catastrophic-risk thresholds. The Center for AI Safety’s well-known extinction-risk statement was signed by figures including Sam Altman, Demis Hassabis, Dario Amodei, Geoffrey Hinton and Yoshua Bengio.
OpenAI: Introducing Superalignment →
Anthropic: Responsible Scaling Policy v3 →
Center for AI Safety: Statement on AI Extinction Risk →
I do not think those warnings should simply be dismissed. I also do not think forecasts should be presented as if every link in the chain has already been demonstrated.
That is the argument I made in AI Made It Possible: If AI Is Going to Kill Us, Show Me the Evidence. Capability is not the same thing as agency. Prediction is not proof. A risk can be serious without being certain.
Reuters reported on September 22 that public concern is substantial: in a Reuters/Ipsos poll, 73% of U.S. adults said AI companies were not doing enough to prevent potentially catastrophic outcomes, while 55% supported slowing development. That does not establish that catastrophic AI is imminent. It does establish that the debate has moved well beyond specialist circles.
The money behind “superintelligence” is not theoretical
The capability may still be debated. The capital race is already real.
Reuters reported this week that U.S. hyperscalers and major semiconductor firms face more than $3 trillion in off-balance-sheet commitments and guarantees tied to the AI buildout. Separate Reuters analysis has shown how rapidly AI capital expenditure is consuming free cash flow across major technology companies.
Reuters analysis on AI obligations and competitive pressure →
Reuters analysis on AI capex and cash flow →
This is why I would be careful with the sentence “trillions have been spent building superintelligence.” That overstates what the evidence can prove.
A more accurate description is that trillions of dollars in investment, financing commitments, infrastructure and market expectations are tied to the race for increasingly capable AI systems.
That distinction matters because “superintelligence” is not only a technical prediction. It has become part of the financial story used to justify an extraordinary buildout of chips, data centers, power, networks and model training.
That connects directly to my earlier reporting on AI infrastructure: AI Is Becoming Infrastructure: The First Trillionaire Signal.
Where transhumanism and life-extension culture fit—and where they do not
There is a real cultural backdrop here, but it needs to be handled carefully.
Some prominent technology investors and entrepreneurs have also funded longevity research and life-extension ventures. That overlap is documented. A 2026 Cambridge University Press essay on “longevity capitalism” describes Silicon Valley investment in these ventures and the way aging is increasingly framed as a technical problem that can be engineered.
Read the Cambridge essay on longevity capitalism →
But that does not prove that Trump’s renaming proposal comes from transhumanist ideology. I have found no evidence supporting that causal claim, and I would not make it.
The relevant point is narrower: we are living through a period in which ideas once treated as science fiction—artificial superintelligence, radical life extension, neural interfaces and large-scale human enhancement—are attracting serious capital and institutional attention.
That makes it even more important to distinguish four things:
- scientific research — legitimate attempts to test a question;
- demonstrated capability — something evidence shows can already be done;
- forecast — a reasoned claim about what may happen;
- belief — a conviction about where technology ought to go.
Those categories should not be collapsed simply because the people discussing them are wealthy, influential or technically sophisticated.
The missing variable is human direction
This is where I think the language problem becomes practical.
The public conversation increasingly swings between two dramatic frames.
AI may become so powerful that humans lose control.
This is the catastrophic-risk frame.
AI is so extraordinary that we should call it “Super Intelligence.”
This is the triumphalist frame.
They sound opposed. But both can push the human being out of the center of the story.
Many of the most immediate AI harms we can already document involve human decisions: who deploys the system, who sets the goal, who grants permissions, who connects it to money or weapons or customer data, who decides how much oversight is enough, who profits, who absorbs the error and who is accountable when something goes wrong.
None of that requires denying that future systems could become far more autonomous.
It simply means we should not let future speculation erase present responsibility.
My concern is not that calling AI “super intelligence” makes the machines more powerful. It is that it may encourage humans to think of ourselves as less important to what happens next.
This is also why I resist using “the AI did it” as an accountability escape hatch. The more capable these systems become, the stronger the human governance around deployment should become—not weaker.
The strongest counterargument to my position
If I am going to criticize inflated terminology, I also have to take the strongest case for it seriously.
Trump may simply be trying to communicate that modern AI is not “fake” intelligence and that its capabilities deserve a stronger name. Some people also use “superhuman” in narrow contexts when AI systems outperform humans at specific tasks such as chess, protein-structure prediction, pattern recognition or selected benchmarks.
That argument is understandable.
But a system can be superhuman at a task without being superintelligent in the way frontier-AI researchers use the term. Precision matters most when the stakes are high.
What should creators take from this?
Do not let vocabulary substitute for understanding.
Suno does not become musically omniscient because a politician changes the category name. ChatGPT does not become an infallible researcher. An image model does not suddenly understand truth, consent or authorship the way a human does. An autonomous agent does not become morally responsible because it can complete more steps without supervision.
Evaluate the actual system in front of you:
- What can it reliably do?
- What does it still get wrong?
- What permissions are you giving it?
- What evidence are you keeping?
- What decisions still require human judgment?
- Who is accountable if the system causes harm?
Those questions are less dramatic than “Will superintelligence kill us?” They are also the questions creators can act on today.
What is confirmed, and what remains uncertain?
Confirmed
- Trump publicly proposed “super intelligence” as the new U.S. government label for AI.
- AI research organizations already use “superintelligence” to mean a future capability level beyond present systems.
- The Trump administration has pursued a growth-first AI strategy.
- The administration has also explicitly required controllability and accountability in national-security AI deployments.
- AI infrastructure financing and capital commitments now reach into the trillions.
Still uncertain
- Whether federal agencies will formally replace “AI” with “super intelligence” across rules, procurement and guidance.
- Whether the terminology will survive beyond the current political moment.
- Whether AGI or ASI will arrive on the timelines predicted by leading labs.
- How much autonomy future systems will actually achieve outside controlled environments.
- Whether current catastrophic-risk forecasts will be borne out by evidence.
The name is not the technology
Donald Trump’s proposal is worth covering because language shapes public expectations. But the larger issue is not whether “super intelligence” catches on.
The real issue is whether we allow a powerful technological and financial system to become easier to anthropomorphize at the exact moment when we need more clarity about human control, incentives, deployment and accountability.
I am not arguing that superintelligence is impossible. I am not arguing that AI risk should be ignored. I am not arguing that every warning from a frontier lab is self-interested.
I am arguing for a simpler standard:
Call demonstrated capability what it is. Call forecasts forecasts. Keep humans accountable for the systems we build, fund, deploy and control.
Powerful tools do not reduce human responsibility. They increase it.
Continue the investigation
AI Made It Possible — Human Capability Campaign
The AI Panic Machine: When Does the Dam Break?
If AI Is Going to Kill Us, Show Me the Evidence
Read the AI Made It Possible series hub
AI Is Becoming Infrastructure: The First Trillionaire Signal
Source and verification notes
This article was researched and published September 22, 2026. Primary and high-quality sources include the White House, NIST, OpenAI, Google DeepMind, Anthropic, the Center for AI Safety, Cambridge University Press and Reuters. Political claims are separated from technical definitions and JR analysis. The article does not claim that Trump’s terminology change has been implemented as a binding government-wide legal definition unless and until a formal directive is identified.
