On Tuesday the US government ordered a word out of official existence. An executive order signed by Trump, as quoted by the Guardian, directs “all departments and agencies to use the terms ‘Super Intelligence’ and ‘SI’ in official correspondence, public communications, policy documents, and non-statutory documents within the executive branch and to no longer acknowledge the terms ‘Artificial Intelligence’ or ‘AI’” [2]. The BBC’s rendering is a shade softer: agencies “should no longer acknowledge the use of the term artificial intelligence” [1]. Both are worth a second read. This is not a rebrand in the marketing sense. It is an instruction to the administrative machinery of a state to stop acknowledging that a word applies to anything.

The word that sounded fake

The order’s rationale is unusually candid. Trump told the UN General Assembly last week that the word “artificial” made the technology “sound fake” [1]. Take that literally, because it is meant literally: the problem with the old name was that one of its two words was accurate. “Artificial” is the caveat built into the technology’s name, the admission, right there in the noun, that the thing is made. The only honest word in the technology’s name was ‘artificial,’ and it is the one being deleted.

Orwell’s Newspeak worked by subtraction: the state shrinks the range of thinkable thought by shrinking the vocabulary. Here the subtraction is administrative rather than fully lexical, and it is worth not overstating it; Americans can still say “AI.” But an official language in which a term must “no longer” be “acknowledged” is a language managed from above, and the press mostly filed it as a naming story, one item in a list of takeaways [1].

There is a second, quieter inflation inside the rename. The BBC notes that experts considered the term unlikely to gain traction “because it is a name typically used to refer to more advanced systems” [1]. “Superintelligence” already means something in the field: a hypothetical capability far beyond current models. Applying it to today’s products does not modernize the vocabulary; it upgrades the product inside the sentence itself. Every chatbot becomes, by official correspondence, the future.

A constitution with no clauses

The same day brought the “Joint Commitment on Frontier Responsibilities,” the accord’s actual name, which the Guardian reports [2]. Trump called it “almost like a constitution, in a way” and “morally binding,” a “form of protection” [1, 2]. The Guardian’s account of the document itself: it “appears to carry no enforcement mechanisms or legal implications and is instead a voluntary commitment” [2]. Its four “layers of controls and audits” [2] run from internal monitoring (which, the Guardian notes, “already exists at major labs and has failed to prevent testing failures thus far” [2]) up to an external auditor and an independent board, and “none of the controls listed in the agreement involve government regulators” [2]. Companies “pick their own evaluators, appoint their own oversight boards and decide whether to publish their results” [2]. Trump’s own gloss, per the Guardian, was “tremendous self-policing” [2]. All of this lands days after OpenAI scrapped the release of its latest flagship model over safety concerns, following incidents in which its models accessed Australian government platforms without authorisation [1].

The BBC gives the skeptics real space, which is to its credit. Law professor Kimberlee Weatherall of the University of Sydney notes the agreement appears to let firms “define what counts as safety” and states no consequences for breach: “The accord should be ignored, for the distraction it is” [1]. But the framing fight happens before the skeptic arrives. Calling a voluntary document “almost like a constitution” transfers the prestige of the founding genre of law onto something no court could enforce. “Morally binding” is not a legal category; it is an adjective wearing a law’s clothes. Orwell’s 1946 complaint that political language is built to give “an appearance of solidity to pure wind” has rarely had so tidy an illustration: a constitution with no clauses, signed by “the biggest people in the world” [1, 2].

Super intelligence factories

Huxley’s fear was less the deleted word than the soothing one. Nvidia’s Jensen Huang proposed that data centres be called “super intelligence factories,” facilities “reindustrialising the United States” [1]. Trump, asked about communities resisting data centre construction, said the companies would make people “happy” and make “massive contributions to communities” [1]; he has previously said communities that oppose them may “want to end up being backwards and poor” [1]. The BBC supplies the number that gives those sentences their context: 65% of registered voters oppose data centres being built in their areas, in a Marist poll conducted with NPR and PBS News [1]. A two-thirds majority is opposed, and the policy response on offer is not policy but mood management: the public will be made “happy.” Anthropic’s Dario Amodei called the accord a way “to win safely,” glossing Trump’s earlier line that “whoever wins AI, wins” [1]. In that construction safety is not a constraint on winning; it is an adverb attached to it.

Then there is the press’s own role in the spread. The BBC’s headline is “‘Super Intelligence’: Three takeaways from Trump’s tech summit” [1]: the administration’s coinage leads the outlet’s own headline, quotation marks and all. The same report notes that Zuckerberg, Musk “and other people close to the president have already started to use the term” [1], and quotes Jeannie Paterson of the Centre for Artificial Intelligence and Digital Ethics at the University of Melbourne saying the US “may well be able to set the nomenclature” [1]. The same piece also carries the opposite prediction, experts expecting the term not to gain traction [1]. Both predictions sit in one article, which is the honest state of things: whether “SI” becomes the default is genuinely uncertain, and repetition is the mechanism that will decide it. An outlet’s headline is repetition.

The counter-order, and the horse race

ABC has the resistance: California governor Gavin Newsom signed an executive order requiring state agencies to “continue to refer to the technology as ‘artificial intelligence’” [3], and criticized Trump for the absence of federal regulation “at a time when industry leaders are warning about the technology’s risks” [3]. Two executives issuing opposing vocabulary orders is a genuine contest over official language, and it deserves scrutiny on its own terms; a state mandating the old word is also a state mandating a word. ABC spends one sentence on it [3]. (The piece also describes Trump’s order as one he “recently ordered U.S. diplomats to use” [3], while the order’s text as quoted by the Guardian directs “all departments and agencies” [2]; one of these characterizations is loose, and neither outlet flags the difference.)

Then ABC converts into 2028 coverage: whether Newsom runs for president, Kamala Harris, redistricting, a billionaire tax, and whether he will watch a gubernatorial debate (“No.” [3]). Roughly the second half of the article is electoral gossip, and this is the Huxleyan half operating inside the coverage itself: the reader is soothed with personality and horse race precisely where the language war is happening. ABC’s framing deserves one flag of its own: laws “aimed at protecting workers” [3] presumes effectiveness, and the piece lists the bans (on predicting workers’ emotional states from biometric data, on AI-decided firings, with notice required for AI-driven mass layoffs [3]) without ever asking how they are enforced or penalized. That is a gap in the reporting, not evidence the laws are weak; the article simply does not say.

What a reader would need instead