Superintelligence Starts at the Table
Updated Sept. 29, 2026
Great technologies do not order themselves. They are put in order when those who build them and those who govern sit at the same table. On Tuesday, in Washington, that is exactly what happened.
Donald Trump lunched in the East Room of the White House with some twenty of the most influential executives in technology, at a gathering organised with Mike Johnson, the Speaker of the House of Representatives. Among them were Jensen Huang of Nvidia; Elon Musk of SpaceX, Tesla and xAI; Mark Zuckerberg of Meta; Sundar Pichai of Google; Jeff Bezos of Amazon; Satya Nadella of Microsoft; Dario Amodei of Anthropic; Lisa Su of AMD; and Greg Brockman, president of OpenAI.
Afterwards, the president announced that he and a broad group of those leaders had signed a voluntary agreement to self-regulate artificial intelligence, with internal and external reviews of the technology. He described it as a "morally binding" document. And he confirmed that the government will officially call AI "Super Intelligence," a change set down that afternoon in an executive order.
I write from Miami, from the side of those who allocate capital for the long term. And from that side, what I saw on Tuesday is good news: the beginning of a structured conversation about the most important technology of our generation.
Coordination before confrontation
For years, the relationship between Washington and Silicon Valley swung between indifference and a test of strength. Today is different. Government and industry recognise, in the same room, that the development of AI requires analysis, oversight and shared responsibility.
Johnson described the document as a declaration of principles by the industry. It is, by definition, a starting point. Its text, its signatories and its mechanisms will become known when it is published, and that will be the natural next step in the process.
Trump added another piece: a ten-person committee to oversee the industry is being considered. It is still an idea under study, but it points to what AI needs: a stable body that accompanies its evolution.
What is already official
The executive order "Inaugurating the Era of Super Intelligence," published by the White House on Sept. 29, directs the executive branch to use the terms "Super Intelligence" and "SI" in place of "Artificial Intelligence" and "AI." In its words, the new name "more appropriately captures the promise, potential, and rapidly advancing capabilities of these technologies."
Federal agencies are to apply it, "to the maximum extent permitted by law," in official correspondence, public communications, websites, reports and policy documents. Regulations, contracts and documents already issued need not change. For now, the legal definition of artificial intelligence applies. Within 60 days, the Assistant to the President for Science and Technology must propose legislative language for a Federal definition of "Super Intelligence."
The president also signed another order, creating America.gov, the unified digital front door to the federal government. It is run by the General Services Administration, the National Design Studio and the Office of Management and Budget, which has 90 days to issue an implementation memorandum. One of its aims is that the super intelligence used on the portal be accurate, reliable and transparent. According to the White House, later this year it will be possible to renew passports and enrol in Medicare there.
These are concrete signals: the name becomes official policy, and AI moves from speeches to public service, with explicit standards of accuracy and transparency.
Last week, Washington and Beijing took a notable step during Xi Jinping's state visit. According to the White House fact sheet, both governments agreed to use the term "super intelligence," created a bilateral dialogue on its risks and benefits — with the next exchange due by November — and established a communication channel for incidents. Axios compared it to the Cold War red telephone.
The comparison is telling. A direct channel for incidents recognises that the technology deserves the care that nuclear deterrence once did. It is institutional maturity.
The industry's vision
The most encouraging part was the tone of the executives themselves.
Zuckerberg said they want to give the American people and their customers confidence that the technology works in the way it was intended. Lisa Su said she was "very encouraged" and spoke of the industry's responsibility to show the power of the technology and ensure that it is very safe. Alex Karp of Palantir said before the lunch that the industry has to take responsibility for the dangers it is aware of.
Huang put the focus on scale: AI is driving the single largest infrastructure buildout in human history, creating an enormous amount of jobs and amounting, in his words, to the largest industrial revolution in history.
Sam Altman of OpenAI told CNBC that they are pacing their progress, which sometimes includes not training a model. Amodei had called this month for a slower pace and dined with Trump on Sunday. OpenAI postponed the launch of a new model on safety grounds.
Speaking at once of responsibility, safety and pace, and of infrastructure and jobs, is exactly the balance one wants to see.
Infrastructure with communities
One commitment from the lunch deserves special attention. According to Trump, the companies committed to supporting the communities where data centres are built, with more resources for schools or measures to lower the cost of energy.
For an investor, this point is central. AI does not live only in the models. It lives in land, in electricity and in the relationship with the neighbours of each facility. An industry that takes on that responsibility builds on firmer ground.
Structure, oversight and a switch
I believe the new era of AI demands going further still, and this lunch opens the door to doing so.
A technology of this reach needs to be supervised and structured. It needs internal and external evaluations, clear rules, bodies that follow its progress and channels to manage incidents. And, in my view, it also needs something any engineer of critical systems recognises: the ability to stop it if necessary. An off switch.
This is not about distrust. It is about design. Power grids, aviation and financial markets work because they have switches, protocols and oversight. Superintelligence deserves the same standard, and coordination between government and industry is the natural place to build it.
What comes next
The calendar offers clear milestones. The details of the accord will become known. The proposed Federal definition of Super Intelligence is due towards the end of November. The next round of the dialogue with China is due by November. The America.gov memorandum will arrive towards the end of December. And the oversight committee remains under evaluation.
Each of those steps is an opportunity to shape the technology that will define the next decade.
Superintelligence will not be built only in laboratories. It will also be built at tables like Tuesday's: with coordination, with analysis and with oversight. It is a good start.
Comments