Musk, Huang and Amodei in Attendance: Trump's "Super Intelligence Luncheon" Gathered the AI Chiefs
When President Donald Trump on Tuesday, September 29, 2026, launched the new AI-powered government site America.gov, it promises free answers to questions about pensions, Social Security and health — with citations to official federal sources.
It did not take long for that promise to face a reality check: the chatbot first answered correctly that it was not election fraud that decided the 2020 presidential election, and then began refusing to answer political questions — while Trump was still delivering his speech at the launch event.
The same day, Trump announced that he plans an executive directive to rebrand "artificial intelligence" as "super intelligence" in all public documents, signed — according to Newsweek — a directive to integrate federal services with America.gov, and had lunch with the top executives of the AI industry — while calling the safety warnings from these very companies a "hoax."
"I will never stifle"
At the launch, Trump framed the technology as something that simply should not be restrained. "This is another example of why I will never stifle the growth of technology that is bigger than the industrial revolution. That's why SI — superintelligence — as they say, is bigger than the industrial revolution, far bigger than the internet, not even close — and we're seeing it, and we can't stifle it. And we're leading by a lot" (ABC News).
He added a pure race frame about winning: "Whoever wins superintelligence wins," he said according to Fox News Digital, adding that there is likely to be no second place — "you get a winner, or you get a loser."
He categorically dismissed the warnings about risk. As early as September 14, he had written in a social media post: "I am the Hoax Buster, and I'm breaking a new hoax right now — that AI is going to take over, swallow up and destroy the world, and that robots will march into our cities and get rid of us all!" (USA Today). In a mid-September post, quoted by ABC News, he called the idea that AI "takes over the world, destroys humanity and everything else that's wrong," a "HOAX" — and compared the warnings to his own impeachments.
His answer to dangerous development paths was not regulation, but law enforcement. "We want people to behave honestly. We want them to be trustworthy, and that's why we have DOJ. That's why we have the FBI," Trump said according to ABC News. On China, he said that in his conversations with Xi Jinping he deliberately would not commit to cooperation or regulation: "We're leading, so why would I do something?" He rejected proposals for joint development with China because the Chinese would then "immediately get access to things we know they don't know."
The table at the "Super Intelligence Luncheon"
The same day, the White House hosted what Fox News called the "White House Super Intelligence Luncheon," with House Speaker Mike Johnson and members of Trump's administration in attendance. According to a seating chart Trump himself shared, he sat in the middle of the long side, flanked directly by Elon Musk (Tesla and SpaceX) and Nvidia's chief executive Jensen Huang. Among the other guests were Google's Sundar Pichai, Meta's Mark Zuckerberg, Microsoft's Satya Nadella, Anthropic's Dario Amodei, OpenAI's president Greg Brockman, Palantir's Alex Karp, Amazon founder Jeff Bezos and investor David Sacks (Fox News Digital).
The White House guest list as reported by ABC News is largely the same, but does not mention Nadella or Sacks. So the sources alone show some disagreement over who actually attended — the difference between a guest list and a seating chart cannot be resolved with what is available. Sam Altman is not listed in either list.
The counterweight to Trump's "hoax" rhetoric came from the table he himself had set. Amodei wrote earlier in the month: "Like many technologies before it, AI brings risk, and because it is such a powerful technology, those risks are serious. They include the risk of losing control of AI systems, misuse of AI for cyberattacks and bioterrorism, and severe economic disruption" (USA Today). Johnson used the word "balance" in an interview with Fox Business: "We want to keep America's leadership over China and every other competitor, but we want to make sure it's done in a safe way. So a little oversight, a little transparency, I think, will go a long way here" (Washington Post via Yahoo News).
What America.gov actually is
The chatbot on America.gov is free and is, according to the White House, intended to "replace" the public websites people normally go to for information about pensions, Social Security and health. It will not store or collect user data, and in 2027 it is to gain features for passport applications, Medicare registration and searches for federal jobs (The Verge).
Newsweek describes an executive directive signed September 29 that orders federal agencies with public services used by more than 100,000 people annually to integrate with America.gov — with exceptions for IRS form submissions and certain national security functions (Newsweek). USA Today, meanwhile, writes only that Trump "plans" a directive on the rebranding to "super intelligence." The sources do not clarify whether this is one or two documents, or to what extent the rebranding order has been signed.
CMS administrator Mehmet Oz defended the solution as a "federated system," in which data is not stored in one place: "We don't want all the information in one link. The real risk we have is that if all your data is in one place ... anyone who hacks it gets everything," he told Newsweek. Trump went further: "It cannot in theory be hacked into. And when they find a way to do it, we'll shut it down, because they always find a way, right?" (Newsweek).
The chatbot's brief career as a source checker
America.gov says its answers are drawn from official information with citations. That worked at first: Asked about the 2020 election, the bot answered, according to the Associated Press: "Official federal sources do not show that widespread fraud changed the outcome of the 2020 presidential election. Isolated election-law violations did take place and have been prosecuted." The bot also contradicted Trump on a third term as president and on the Pentagon name, AP reports (AP via TwinCities).
Then, while Trump was still speaking at the launch, the site began changing its answers. Asked again about the 2020 election, it said it does not answer "political questions" — it gave the same answer when asked whether Trump had been impeached or convicted of serious crimes (AP). PBS confirms the sequence: after first stating that Biden won in 2020, the bot subsequently refused to answer, citing that it does not respond to political inquiries (PBS News).
How the change happened is unknown. The mechanism — human intervention, changes to the setup, or the model's own behavior — is not resolved in any of what is available, and for now must stand as an open question.
Which model? Even that is unclear
A seemingly simple technical question — what powers America.gov — is in fact disputed. PBS writes that the site does not disclose its platform: asked about its own systems, the bot answered that it "does not publicize the name of a specific AI model," but "uses AI to search official public sources and explain them with citations."
The Washington Post (via Yahoo News), however, quotes Joe Gebbia, the US chief design officer and the architect behind America.gov, saying the site is a partnership with Google's Gemini and Grok, and describes it as a "concierge to navigate government." The sources thus conflict: PBS says the platform has not been disclosed, while Gebbia, according to WaPo, names Gemini and Grok. Both accounts are presented here as they stand — neither can be confirmed as the final answer.
What remains
Several things are unresolved after launch day. That includes the signature status and exact contents of the executive directives — particularly the rebranding to "super intelligence" — the differences between the guest list and the seating chart, and not least what America.gov actually answers now, after the political gates closed during the launch ceremony. It is also an open question whether the 100,000-user threshold in the integration directive and the exceptions for the IRS and national security will hold in practice. The only thing that can be stated with certainty is the timeline: all of this happened on the same day — and the chatbot began withdrawing from politics before the speech was over.