Can ‘super intelligence’ and a non-binding safety pact solve AI’s image problem?

President Trump hosted many of the biggest names in AI this week, where executives signed a “Joint Commitment on Frontier Responsibilities” and the administration announced, by executive order, that official U.S. representatives should now say super intelligence instead of artificial intelligence. Over the weekend Trump followed that with a new Super Intelligence Force.

On TechCrunch’s Equity podcast, Anthony Ha, Kirsten Korosec and Sean O’Kane were skeptical the pact will amount to much. It is voluntary and legally non-binding; Trump has described it as morally binding. As O’Kane noted, the signed document even misspelled the United States.

Why it matters: this is the federal government’s main safety gesture toward frontier AI right now, and it arrived just after the President had been dismissing the need for any constraints at all. If the pact is the policy, there is no enforcement. If it isn’t, we still don’t know what is.

We’ll take the charitable read first. Getting Zuckerberg, Bezos, Musk, Amodei and the rest of the industry’s power brokers in one room to sign anything about frontier responsibility is not nothing. Only a week or two earlier, as the hosts recount it, the President was waving off anyone who warned about AI danger. Now there’s a written commitment with CEO signatures on it. Norms often start as voluntary promises before they harden into rules, and a public pledge at least gives the press and the public something to hold companies to.

But we can’t pretend the pledge is a safety regime. There is no audit, no penalty, no independent check. “Morally binding” is a phrase that asks for trust at exactly the moment trust is thinnest — in the same stretch that brought a frontier model cancelled for failing internal tests and growing public fear about jobs. A rushed document with a typo in the country’s name is not the artifact of a careful process.

Korosec’s point is the one we keep coming back to: the meeting looks like a rebranding exercise. AI is the scary thing that takes jobs; super intelligence is the patriotic thing the country leads in. We think that’s a risky trade. Changing the word doesn’t change what the systems do, and people whose work is being reshaped will notice that the vocabulary moved faster than the protections. If the label becomes a way to wave off hard questions, it will deepen the image problem, not solve it.

The Anthropic thread is worth watching. Amodei went from the administration’s doghouse to a 10 p.m. dinner with Trump to this luncheon within days. Yet the podcast notes the Defense Department still treats the company as a national security risk, with Emil Michael attacking it on X, while the Treasury Secretary has been friendly. Different parts of the government want different things from the same lab. That kind of split makes a voluntary pact even weaker, because there’s no single government position it answers to.

We’re for useful AI and for the U.S. building it well. That case is stronger, not weaker, with real oversight behind it. Where this leads if it scales: if safety becomes a photo op plus a rename, the first serious incident will land on a public that was told everything was handled — and the backlash will hit good tools along with bad ones.

The executive order on terminology, the Joint Commitment signing and the Super Intelligence Force announcement all came within days of each other. The hosts suggested the pact came together quickly, after a stretch in which the administration had opposed regulation and Anthropic had been in open conflict with the Defense Department over conditions on how its technology could be used.