Washington Took the Safety Movement's Word and Left the Demands Behind

On August 11, "superintelligence" was the word of the people who wanted the technology governed. An open letter signed by 1,367 researchers and executives from OpenAI, Anthropic, and Google DeepMind warned of a catastrophic "intelligence explosion" and called on the U.S. government to lead an international effort built on three things: a formal licensing process, registration of AI systems, and enforceable international agreements ^[1]. Two of the most prominent signers, Dario Amodei and Demis Hassabis, offered to halt their own development if competitors agreed. The word was a demand, not a brand. Six weeks later it reached the podium of the United Nations General Assembly, carried by a government that wanted none of what it had asked for. On September 22, President Trump announced the United States would replace "artificial intelligence" with "super intelligence" in all official documents — in the same speech that rejected "any attempt to construct a globalist scheme to control" the technology and waved off warnings of existential risk, the fear that the technology could one day end or radically reshape human life, as a hoax ^[2].

The United States also totally rejects any attempt to construct a globalist scheme to control for the artificial intelligence being spoken of so much now, here and after officially called super intelligence, changing the name, and that the use of the word artificial makes intelligence fake. — Donald Trump

To be plain about what this is: nothing on the record establishes that the word was taken from the researchers as a deliberate confiscation. The president's stated reasons are aesthetic — "artificial" makes intelligence sound fake — and strategic, a lead to be won over China ^[3]. What the record does establish is the sequence: the safety movement's word was ordered into the government's documents, and the demands it had been coined to carry were refused in the same breath. Since then, almost everything attached to the name has been voluntary. There is an accord, signed by Google, Meta, Nvidia, xAI, Anthropic, and OpenAI, that its own champions call "morally binding" — a self-policing promise whose only trace in the archive is the story of its signing, with no penalty or enforcement mechanism following it ^[4]. There is a czar. Jay Clayton, tapped to lead the effort while keeping his intelligence portfolio, has put the name's governance logic into a single sentence.

AI is the next Industrial Revolution, or Internet, but will be even larger and more impactful, possibly as much as 25% of our Country’s GDP. — Donald Trump

He argues that the Justice Department and the Federal Trade Commission — the agencies that already exist — are the right bodies to police the companies, rather than any new regulatory framework ^[5]. In this reading, national-security status is a reason to build nothing new. On the opened record, that national-security logic has met something that could actually bind exactly once — and it ran against safety. Earlier this year, months before the word reached the podium, the Pentagon delivered Anthropic an ultimatum: strip Claude's guardrails against mass domestic surveillance and fully autonomous weapons, or lose the military's business. The lab refused, and the government flagged it as a national-security supply-chain risk — a government marker that a company is a security threat in the technology it supplies — and President Trump ordered federal agencies to stop using its products ^[6].

These actions are unprecedented and unlawful. The constitution does not allow the government to wield its enormous power to punish a company for its protected speech. — Anthropic

The name has otherwise been adopted by almost nothing with power to bind. Days later, at the same assembly, foreign ministers still said "AI" — Germany, Singapore, and Malaysia among them, none taking up the new word ^[7]. No court has adopted it, no state has adopted it, and its clearest adoption outside Washington is commercial, in the hands of Slovenia's .si domain speculators, where registrations jumped from under 2,000 in August to 44,000 in September ^[3]. On the record, the rebrand has deflected exactly zero legal actions against a company. The one bill written in the new vocabulary turns it back against the industry. Two days after the rebrand, Bernie Sanders and Greg Casar introduced the Ban Artificial Superintelligence Act, which would pause systems trained past a threshold of computing power, attach 20-year prison terms, and seize the charter of any company that builds one ^[8].

Do we really want to develop a super intelligence that when it becomes smarter than human beings could act independently of human control? I don’t think we do. — Bernie Sanders

The new word's first legislative use, in other words, sends it back to its original job. In August, the letter asked for three things: licensing, registration, enforceable international agreements. Count them on the opened record and none has been enacted. The word itself, meanwhile, was ordered into all official documents by the same government that refused what it asked for ^[2].