I was halfway through a coffee when the wording landed on my desk, and I actually laughed out loud. Not because the stakes are small. Because two powerful offices, a day apart, decided the most urgent move was to argue about what we call the machines. One side wants the country to say Super Intelligence. The other told its own agencies to keep saying artificial intelligence, full stop. If you work with technology, hire people, or just try to understand the next round of rules, that little fight is not a sideshow. It is the opening line of a much larger argument about power, safety, and who gets to write the dictionary.
Names are sticky. Once a phrase sits in a statute, a contract, or a job posting, it does not peel off cleanly. I have watched teams spend months renaming a product only to discover the old term still lives in every vendor agreement and every training slide. So when a federal directive and a state directive collide over a two-letter acronym, the practical question is not which phrase sounds cooler. It is which phrase your lawyer, your HR lead, and your software vendor are allowed to print.
A One-Day Clash Over What the Machines Are Called
On September 29, a presidential order told the federal executive branch, to the maximum extent the law allows, to swap artificial intelligence and AI for Super Intelligence and SI. Official letters, websites, reports, policy papers, and other non-statutory documents were in scope. Older rules, contracts, grants, and historical files did not have to be rewritten. A fact sheet added that the science adviser should propose a federal definition of the new phrase within 60 days.
The public explanation was blunt. The word artificial, the argument went, suggests something fake or false, and the technology is neither. Alternatives such as superior intelligence and supreme intelligence were considered and set aside, partly to avoid a collision with the name of the high court. The claim that followed was competitive: only the United States and China are truly in this race, the United States is ahead, and whoever wins the field wins the era.
California answered the next day. An executive order under the governor’s authority told agencies and departments to keep referring to the technology as artificial intelligence and AI, even if Washington uses a different label, unless the law itself requires something else. The text did not name the president. The accompanying statement did. The line that traveled farthest was a jab: super intelligence is clearly not coming from the White House, which is why California keeps leading.
Changing a name does not distract a person of ordinary judgment from noticing a failure to act on documented safety and security risks.
Paraphrase of the state order’s core argument
That sentence is the whole dispute in one breath. One office says the old word undersells a real capability. The other says a rebrand is a distraction from guardrails. Both can be partly right. A label can be clumsy and still be the term every engineer, insurer, and judge already understands. A new label can sound ambitious and still leave the hard questions untouched.
Why the Acronym Fight Is Not Cosmetic
Perhaps the most interesting aspect is how fast language becomes infrastructure. Procurement templates, grant portals, audit checklists, and public dashboards all key off phrases. Tell one layer of government to refuse the old term and another layer to refuse the new one, and you get a translation tax. Staff will keep a glossary in a drawer. Vendors will ship two versions of the same slide. Reporters will spend the first paragraph explaining the nickname instead of the rule.
I have found that the teams who suffer first are not the speechwriters. They are the people who file forms. A clinic updating a consent notice. A university lab describing a model in a grant. A city IT shop labeling a chatbot on a benefits site. If the federal document says SI and the state document says AI, someone has to decide which box to tick before the deadline. That decision is boring. It is also where policy becomes real.
There is a second cost, and it is cultural. Workers hear the new phrase and wonder whether the goalpost moved. Super intelligence sounds like a finish line, not a tool. Artificial intelligence sounds like software with a marketing problem. Neither phrase is precise. Both steer attention. If you manage a team, the phrase your boss repeats will shape which risks they escalate and which ones they shrug off.
What the Federal Side Actually Asked For
Read narrowly, the federal instruction is a style rule for the executive branch. It covers correspondence and public communications. It tells agencies not to acknowledge the old usage in applicable settings. It does not rewrite statutes that already say artificial intelligence. Congress still owns the code. Courts still read the words on the page. A press office can prefer SI on a fact sheet and still have to quote a statute that says AI.
The definition task matters more than the nickname. A 60-day proposal for what Super Intelligence means could be narrow, technical, and useful, or it could be a slogan with a footnote. Until that definition exists, SI is a mood. It signals speed, national competition, and impatience with the word fake. It does not yet tell an inspector what system counts, what capability threshold applies, or what a company must disclose.
- Official letters, websites, and reports are pushed toward the new term.
- Older regulations, contracts, and grants can keep their original wording.
- A science adviser is asked to propose a definition on a short clock.
- The order frames the United States and China as the only serious competitors.
- Statutory language is left alone, which limits how far a style rule can travel.
That last point is easy to miss in the noise. A president can tell his own agencies how to talk. He cannot, by pen alone, force a state, a private employer, or a foreign regulator to adopt the same dialect. California just demonstrated the limit in public.
What California Told Its Own Agencies
The state order is narrower than the headlines suggest, and also sharper. It binds agencies and departments under the governor’s authority. It creates a default: say artificial intelligence, say AI, unless the law points elsewhere. It treats the federal rebrand as something to be noticed and then ignored in ordinary state writing. The tone is dismissive of renaming as a substitute for action on security and safety risks that, in the order’s view, are already well documented.
The same day, the governor signed 13 bills touching workplaces, health care, digital watermarks, deepfakes and digital replicas, legal work, and firms that synthesize genes. His office called the workplace package a national first. Labor leaders praised four of the measures as a way to put guardrails on the tools and put workers first. The pairing was deliberate. Terminology on one sheet. Statutes on the other.
In my experience, that pairing is how states answer a federal message they dislike. They do not only issue a counter-slogan. They pass something a compliance officer can implement on Monday. Whether those bills are wise is a separate argument. The political move is clear: we will keep the old name, and we will attach new duties to it.
Workplace Rules That Outlast a Nickname
The employment measures are the part a manager should actually read. They restrict employers from relying only on an automated system when they discipline or fire someone. They require disclosure when an automated system is a cause of a mass layoff, a relocation, or a termination. They ban certain surveillance tools in bathrooms. None of that depends on whether the press release says AI or SI. A dismissal is a dismissal. A camera in a restroom is a camera in a restroom.
Labor’s argument is straightforward. Workers asked the state to lead on workplace use, and the bills are a start. The governor’s argument is that the technology should widen opportunity rather than come at the expense of workers and families. Critics will say the rules add cost, invite lawsuits, and freeze tools that smaller firms need. Supporters will say a human still has to own a firing decision, and people deserve to know when software helped empty a floor.
I lean toward the disclosure piece, with a caveat. Telling people that a system played a role in a layoff is basic respect. Pretending the system made the call alone is how companies dodge accountability. The harder design problem is the middle: most decisions are mixed. A model ranks resumes. A manager picks from the short list. A finance lead sets the headcount. If the law says “only” automated decisions are barred, clever process maps will keep a human signature on the last line. Good policy has to anticipate that shrug.
| Workplace issue | What the new approach stresses | What still sits with the employer |
| Discipline or firing | Do not rely on the system alone | A person who can explain the decision |
| Mass layoff or relocation | Disclose when a system is a cause | Timing, notice, and the business reason |
| Bathroom surveillance | Specified tools are off limits | Any other monitoring still needs a policy |
| Everyday scoring tools | Not fully settled by a nickname change | Vendor contracts and audit logs |
Notice what the table does not say. It does not say the federal label cancels the state duty. A company with staff in California will follow California employment law whether the White House style guide prefers SI. Multi-state employers already live in that split. This episode just makes the split louder.
Health Care, Replicas, and the Quiet Bills
The workplace cluster got the quotes. The rest of the 13 bills may age differently. Health care uses of these systems touch consent, error, and liability in a way a marketing chatbot does not. Digital watermarks and rules on deepfakes and digital replicas speak to a problem everyone has already met: a face or a voice that is not quite the person. Legal-work provisions nudge a profession that bills by the hour and is suddenly staring at drafts that arrive in seconds. Gene-synthesis companies sit at a stranger intersection of software screening and physical biology.
I am not going to pretend each bill is equally mature. Some of these domains have years of adjacent law. Others are being written while the capability is still moving. The useful habit is to ask, for each one, what behavior changes on a Tuesday. Does a clinic have to label an output? Does a platform have to watermark a synthetic clip? Does a synthesis firm have to screen an order? If the answer is a concrete duty, the bill is policy. If the answer is a press sentence, it is atmosphere.
Atmosphere still matters. A state that signs a stack of bills on the same day it rejects a federal nickname is telling companies where the next enforcement energy will go. You can dislike the politics and still update your compliance calendar.
Safety Talk Versus Capability Talk
Here is the tension I keep coming back to. One camp talks about promise, potential, and rapidly advancing capability, and wants a name that matches that climb. The other camp talks about emerging security and safety risks, and treats a rename as an impotent substitute for action. Both sentences can be printed on the same week without either side defining the risk in numbers.
What counts as a documented risk? Model-assisted fraud. Synthetic media used in scams. Biased screening in hiring. Leakage of sensitive data into a vendor tool. Fragile automation in infrastructure. Over-trust in a clinical suggestion. None of those require a science-fiction label. They require logs, liability, and someone who can be called. A grander name does not remove them. A sarcastic counter-order does not measure them either.
A sharper name is not a safety case, and a safety case is not a growth strategy. Serious offices have to hold both without using one as an excuse to skip the other.
Recent policy debates, across several countries, keep circling the same trio: capability, misuse, and institutional readiness. The United States conversation often adds a fourth item, industrial competition with China. That fourth item explains the urge to sound maximal. It also explains why a state with a large technology workforce wants to sound like the adult in the room on labor and consumer harm. The rhetoric is opposed. The underlying map is shared. Both sides know the tools are already in payroll systems, hospitals, studios, and laboratories.
How Companies Should Read the Noise
If I were briefing a general counsel this week, I would separate speech from duty. Speech is what the website says. Duty is what the statute requires. Update the federal-facing style sheet if you draft documents for an agency that now prefers SI. Do not scrub historical contracts for sport. Keep California filings in the vocabulary California agencies will accept. And do not let the acronym debate delay the workplace controls you already know are coming.
- Map which documents are federal communications and which are state filings.
- Keep a two-column glossary so staff are not guessing in email.
- Review hiring, firing, and layoff workflows for sole reliance on a model.
- Check surveillance policies against location-based bans, including restrooms.
- Ask vendors how they will label outputs if watermark rules tighten.
- Treat the 60-day federal definition as a calendar item, not a rumor.
That list is deliberately dull. Dull is how you stay out of the worst version of this story. The dramatic version is a CEO on a stage saying super intelligence while a state investigator asks why a restroom camera was still online. You do not want to be the bridge between those two sentences.
Language, Law, and the People Who Never Voted on the Acronym
There is a democratic itch in this episode that is easy to mock and hard to dismiss. Voters did not cast a ballot on SI versus AI. They elected people who now treat vocabulary as an instrument. That is normal in government. It is also why the instrument should stay tethered to statutes the legislature actually passed. When executive orders become the main arena, the public gets a feud and a thin record.
California’s move has the same shape from the other direction. A governor can instruct his agencies. He can sign bills the legislature sent him. He cannot, by statement, prove that his state leads a global technical race. Leading is a claim about labs, capital, talent, and energy supply, not about who wins the press cycle on a Wednesday. I would rather see both offices publish the boring metrics: incident reports, audit pass rates, grant outcomes, grid capacity for new data centers. Nicknames do not generate electricity.
Still, do not underestimate pride. The line about super intelligence not coming from the White House was built to travel. The line about artificial meaning fake was built to travel too. Once those lines are out, staff inherit them. Career officials will soften them in guidance. Campaign accounts will sharpen them. The rest of us get to live in the gap.
A Practical Glossary for the Next Quarter
Until a federal definition lands, treat the phrases as audiences, not as physics. Use the term your reader is required to use. Explain the other term once, then get back to the system. If you write for the public, plain description beats either brand: software that generates text, ranks applicants, reads a scan, or imitates a voice. People understand those verbs. They do not need a coronation.
Working split for mixed audiences: Federal executive communications -> follow the current federal style California agency filings -> artificial intelligence / AI Statutes and old contracts -> leave the original words Internal engineering docs -> describe the system, then the label Worker notices -> behavior and rights, not slogans
That split will annoy brand teams. Good. Brand teams are not the ones who get deposed. The person who sent the termination email is. Write for that person.
Competition Talk and the Missing Middle
The federal remarks framed the contest as a two-country race and said the United States is leading by a wide margin. Whether that margin is real depends on what you count: frontier models, chips, power, talent, open research, or deployed systems in factories and clinics. A single sentence cannot carry all of those. China is a serious peer in several of them. So are clusters of firms and labs that do not fit a flag. Reducing the field to two capitals is a speech tactic. It is a weak map.
California’s counter-claim, that it continues to lead, is also a speech tactic. The state houses an extraordinary share of the companies and researchers. It also houses the housing costs, the power constraints, and the political fights that push some projects elsewhere. Leadership that cannot keep a data center permitted is a press-release leadership. I say that as someone who thinks concentration of talent is a real advantage, not a myth. Advantage is not the same as a permanent crown.
The missing middle is everyone who uses the tools without building them. A logistics firm in the Central Valley. A hospital group. A union local. A community college rewriting its academic-integrity rules. They do not win or lose “super intelligence.” They inherit the defaults. The naming war reaches them as a confusing memo. The bills reach them as a training requirement. Guess which one changes a shift schedule.
What a Definition Would Need to Be Useful
If the science adviser returns in 60 days with a definition, the useful version will be boring on purpose. It will say which systems are covered, by capability or by use. It will say what is out of scope. It will not try to win a branding war inside the definition itself. A definition that exists mainly to retire the word artificial will age poorly the moment the next office wants a different adjective.
Technical communities already have a mess of overlapping terms: machine learning, foundation models, generative systems, automated decision systems, automated employment decision tools. Regulators in several states have started to define the employment slice with more care than the umbrella phrase. That is the right instinct. An umbrella is for speeches. A definition that triggers a duty should be cut to the duty.
Would I bet on a clean definition in two months? Cautiously, no. The incentive is to produce something quotable. The work of scoping exemptions, open-source edge cases, and research use takes longer than a news cycle. If the document is thin, agencies will improvise. Improvisation is how you get five regional meanings of the same two letters.
Workers, Families, and the Opportunity Line
The governor said the technology should expand opportunity rather than arrive at the expense of workers and families. That sentence is popular because it is incomplete. Opportunity for whom? A tool that drafts routine memos can free a skilled employee and erase an entry-level role. A diagnostic aid can shorten a wait and also deskill a junior clinician if nobody protects the training path. Families feel both effects: a faster benefit decision, and a parent whose shift disappeared because a scheduling model optimized them out.
Guardrails do not automatically create opportunity. They can stop the worst uses and still leave the wage question untouched. I would rather hear officials talk about apprenticeships, task redesign, and who captures the productivity gain than hear another riff on the adjective. The bills on disclosure and human involvement are a floor. They are not a labor-market strategy.
Union leaders called the moment a beginning. That is the honest frame. A beginning can stall. It can also spread if other states copy the employment pieces and ignore the nickname war. Copying is how American regulation often actually moves. Not by a single national label. By a stack of state duties that national firms eventually standardize around, because maintaining fifty versions of a firing protocol is worse than maintaining one strict version.
Media, Markets, and the Attention Trap
Markets will try to trade this. They should be careful. A style order is not a chip export rule. A state employment bill is not a ban on model training. The price of a software vendor can still jump on a headline, because headlines are how a lot of fast money reads policy. The slower money will ask which revenue line is exposed to a disclosure duty, a watermark duty, or a health-care liability theory.
There is an attention trap for readers, too. The insult is memorable. The scope clause is not. “To the maximum extent permitted by law” is the phrase that keeps the federal order from pretending to amend statutes. “Unless inconsistent with the law” does similar work on the state side. If you only remember the joke about the White House, you will misread what either document compels.
I keep a simple test for these episodes. Can I explain the operative sentence to a friend who does not follow politics, without the insult? If I cannot, I do not understand it yet. Operative sentence one: federal agencies should say Super Intelligence in their own communications where the law allows. Operative sentence two: California agencies should keep saying artificial intelligence unless the law says otherwise. Operative sentence three: a stack of new state duties now attaches to workplace and related uses. Everything else is color.
Scenes From the Week the Label Split
Picture a communications director in a federal agency on the morning after the order. The website still says AI in the navigation. The new guidance says not to acknowledge that usage in applicable settings. Does she redirect the URL today, or wait for the definition? Picture her counterpart in Sacramento, told to hold the old term even if Washington moves. She is not confused about her instruction. She is confused about the joint webinar they were going to host.
Picture a hospital counsel. Half the guidance she receives from a federal program office may start saying SI. The state health bill she must implement still says artificial intelligence. Her consent form has 400 words and a sixth-grade reading target. She does not have room for a terminology essay. She will pick the phrase the patient already knows, then footnote the rest. That is not defiance. That is triage.
Picture a warehouse supervisor told that a scheduling system contributed to a reduction in shifts. Under the new disclosure idea, someone has to say so. He did not choose the model. He chose who stayed on the Friday roster after the model ranked availability. If the company writes that the system “caused” the cut, he becomes the face of a sentence he did not draft. These are the rooms where the order and the bills eventually sit down.
What I Would Watch in the Next Two Months
First, the definition. If it arrives with examples, exemptions, and a clear link to existing statutory terms, it will matter. If it arrives as a paragraph of ambition, agencies will keep improvising. Second, agency style guides. The gap between an order and a revised manual is where staff make quiet exceptions. Third, California’s implementing memos for the workplace bills. Statutes announce. Memos tell you the form to file. Fourth, whether other states copy the employment pieces or the rhetorical posture. Copying the duties would change the national floor. Copying the insult would only extend the tour.
Fifth, industry self-description. Some firms will adopt SI in federal meetings and AI everywhere else, the way contractors already switch dialects between rooms. Others will refuse the new term on principle or on brand grounds. Neither choice tells you whether their safety work is serious. Ask for the incident log, not the adjective.
- A definition with examples beats a definition with adjectives.
- Style manuals will show how strictly agencies intend to comply.
- State implementing memos will matter more than signing-day quotes.
- Other states may copy duties without copying the nickname fight.
- Vendor contracts will lag both orders unless someone forces a refresh.
A Note on Insults and Governing
Both sides reached for a slight. One called the old word terrible and fake-sounding. The other implied the rival office could not produce the thing it was renaming. Insults are effective at gathering an audience. They are bad at specifying a control. I do not need officials to be dull. I need the operative rules to be findable after the clip ends.
There is also a risk of crying wolf on language itself. If every administration renames the same systems, the public learns that vocabulary is partisan merchandise. Then, when a term really does need to change because it misleads, nobody trusts the edit. Artificial was always a clumsy word. Plenty of researchers have said so for years, without a presidential order. The clumsy word survived because it was shared. Shared is a feature in law.
Could a better phrase exist? Sure. Automated decision support. Generative systems. Machine learning systems. Super intelligence, if someone eventually defines a threshold that the words can bear. Until then, switching the sign on the door and daring the other party to keep the old sign is politics. Politics is allowed. It is not a substitute for the bathroom rule, the firing rule, or the watermark rule.
Readers Who Just Want a Straight Account
So here it is, without the flourish. A president directed his executive branch to prefer Super Intelligence and SI over artificial intelligence and AI in communications and similar documents, as far as the law permits, and asked for a definition within 60 days. He argued the old word implies fakery and that the capability is real, and he framed the contest as one the United States is winning against China. A day later, California’s governor ordered his agencies to keep the old terms unless the law conflicts, argued that renaming does not fix safety failures, and signed 13 bills that include first-in-the-nation style workplace limits on sole reliance, layoff disclosure, and certain surveillance.
You can think the federal rename is overdue clarity. You can think it is branding. You can think the state response is sober governance. You can think it is a campaign line with statutes attached. What you should not think is that the machines paused while the adults argued about the label. They did not. Payroll systems ran. Clinics ran. Studios ran. The argument changed the caption, not the current.
Holding Two Ideas Without Picking a Team
Idea one: words shape budgets. If leaders say the prize is super intelligence, money flows toward frontier capability and national competition. Idea two: words can hide gaps. If leaders spend a news cycle on a synonym, oversight of ordinary harms waits. A grown-up policy can fund capability and still ban a camera in a bathroom. Those are not contradictory projects. They are different floors of the same building.
I have found that people who work closest to the systems are the least impressed by either slogan. They want to know the evaluation set, the access control, the appeal path when a score is wrong, and who gets called at 2 a.m. when a tool misfires. Give them that, and they will call it whatever the form requires. Starve them of that, and no acronym will save the deployment.
Maybe that is the quiet moral of the week. The offices fought over a name because names are cheap and visible. The expensive work is the audit, the training, the energy, the appeal, and the decision to keep a human in the loop when someone’s job is on the line. California put some of that expensive work into bills and then mocked the rename. Washington renamed the field and pointed at a race. The rest of the country gets both messages, and a glossary.
Plain test before you forward the memo:
Does this sentence create a duty, or only a preference?
If it is a duty, cite the bill or the contract.
If it is a preference, say whose style guide you are following.
If you cannot tell, do not send it yet.Use that test on the next all-staff note. It will save you a round of confused replies, and it will keep the feud from leaking into a procedure that was supposed to protect someone.
Where This Leaves the Public Conversation
The public is allowed to find the episode absurd and still take the bills seriously. Absurdity is not a reason to skip the fine print. It is a reason to demand that fine print be written in words that survive the next election. Artificial intelligence has survived several hype cycles precisely because it was nobody’s campaign property. Super Intelligence might become a durable technical term, or it might become a dated slogan. California’s insistence on the old phrase might look principled, or it might look petty, depending on whether the safety work is real.
I will judge both by the follow-through. A definition that agencies can apply. Memos that employers can follow without a law firm on retainer. Incident reporting that the public can see. Power and permitting decisions that match the rhetoric about leadership. If those arrive, the nickname will fade into a footnote. If they do not, we will still be arguing about vowels while the systems keep scoring, drafting, and sorting.
Until then, say the accurate thing in the room you are in. Call it artificial intelligence when the form says so. Call it Super Intelligence when that office requires it. Describe the actual function either way. And when someone tells you the rename itself is the strategy, ask what happens to the worker, the patient, and the record when the system is wrong. That question does not fit on a sign. It is still the one worth staying for.
