Remember when “the cloud” arrived and suddenly every company you’d ever given your email address to was telling you their servers had become weather? Nothing about the machines changed. The word changed. And for a while, the word did a lot of work: it made storing your family photos on someone else’s hard drive sound light, airy, and obviously fine.
I thought about that this week, because the AI industry has just tried a similar trick, except bigger and with a signing ceremony.
What actually happened
Two things, and they’re easy to mix up.
- President Trump signed an executive order replacing the term “artificial intelligence” with “super intelligence,” or “SI,” across US government communications.
- Trump and the leaders of six of the top US AI companies signed a document called the White House Accord on Super Intelligence, which Trump described as “morally binding.”
One is a vocabulary change. The other is a safety agreement. Neither of them is a law.
On TechCrunch’s Equity podcast, Kirsten Korosec, Sean O’Kane, and their co-host talked through what was driving the meeting and what it might actually do to the industry. The read was skeptical across the board. The Council on Foreign Relations put it more bluntly in a headline, calling the pact toothless while arguing there’s still a path forward.
“Morally binding” is not a category
I want to sit with that phrase for a second, because if you’re not steeped in policy language, it can sound stronger than it is.
Legally binding means a court can be involved. There are obligations, and there are consequences when you fail to meet them. Morally binding means everyone agrees it would be bad to break the promise. That’s it. The accord is voluntary, and reporting on it describes it as deeply non-binding.
If you’ve ever been in a group chat where six people agreed to split the dinner bill evenly and then one person “forgot,” you already understand the enforcement model here.
The two gaps that matter
When I’m trying to judge whether an AI commitment means anything, I look for two things. This accord is reported to be missing both.
Enforceability. Who checks? What happens if a company doesn’t follow through? Without an answer, a safety commitment is a press release with signatures on it.
Clear lines of responsibility. If an AI system causes harm, who owns that outcome? The company that built the model? The company that deployed it inside a product? The agency that bought it? The person who typed the prompt? Vague responsibility is how harm ends up belonging to nobody, which in practice means it belongs to the person it happened to.
Those aren’t academic concerns for non-technical readers. They’re the whole thing. Most of us will never audit a model. What we can reasonably expect is a named party who answers for it.
About that typo
The agreement misspelled the United States.
I’m not going to pretend a typo is a scandal. Documents get typos. But reporting flagged it alongside the fact that the accord appears to have been assembled quickly, and together those details tell you something about priorities. A document that was carefully negotiated over months, with lawyers arguing over the meaning of individual verbs, tends not to misspell the country hosting the ceremony. A document produced to be signed on a particular Tuesday might.
The typo isn’t the problem. It’s a hint about the process that produced the thing.
So does any of it help AI’s image problem?
Here’s my honest take as someone whose whole job is explaining this stuff to people who did not ask to have opinions about model safety.
The gesture is real. Getting a sitting president and the heads of six major AI companies into one room, on the record, agreeing that this technology needs guardrails is not nothing. It sets a reference point. It makes future refusals to cooperate more conspicuous. Commentators who doubt the pact’s impact have still treated the meeting itself as significant, and that’s a fair distinction.
But an image problem and a trust problem are different problems, and only one of them can be solved with naming. Calling it “super intelligence” doesn’t make a system more accurate, more private, or more accountable. If anything, the new label is a more flattering description of the same software — and flattery is not a safety feature.
What would move me? Specific commitments with specific deadlines. A named body that can check the work. A clear answer to “who is responsible when this fails.” Those are the things worth watching for in whatever comes next, and they’re things you can evaluate without a technical background. You just read the document and ask who’s on the hook.
For now, we have a new word, a handshake, and a promise that everyone has agreed would be bad to break. Hold onto your skepticism. It’s still the most useful tool you’ve got.
đź•’ Published: