Elon Musk is suing Sam Altman. This week they agreed about something, and it was this: the AI industry should slow down the most dangerous part of what it is building.
The essay that set it off came from Anthropic’s Dario Amodei. It is called “We Must Pace the Frontier.” Three days later the industry had split in a way that does not follow any of the usual lines.
What he is asking for
Three things.
- Outside evaluators inside the labs. Independent people with the same access a staff engineer gets, not a report handed over after the fact.
- Industry-wide safety standards. Amodei admits this needs an antitrust waiver from Washington to even be legal, because competitors coordinating is normally the thing the government stops.
- Government-backed international limits on the most dangerous capabilities, eventually.
It is not a pause. Nothing in it stops anybody from shipping products.
Who lined up where
For it: Sam Altman, who Amodei left OpenAI to get away from, wrote “I agree with Dario that we need to pace the frontier.” Elon Musk, who is actively suing Altman, said three words: “Dario is right.” Google DeepMind’s Demis Hassabis agreed too.
Against it: Mark Zuckerberg says labs already have enough reason to build safely on their own. Jensen Huang, on stage with Amodei and Altman, said labs should pace themselves without being made to. We covered Huang rejecting Amodei’s earlier call for regulation last week, and his 0% odds on AI ending the world.
Trump waved it off as talk, framing it as an arms race with China that America has to win.
The BeezLoop Take
Start with the part that should make everybody pause. The four men who agree here run the four labs that are furthest ahead. The rule they are proposing would apply to the frontier, which is the place they already occupy, and it would be enforced by standards they would help write.
That does not make them wrong. It does mean the agreement between Musk and Altman is less surprising than it looks. Two people in a lawsuit finding common ground on a policy that locks in their position is not a miracle of conscience.
The antitrust waiver is the tell, and Amodei deserves credit for saying it out loud instead of burying it. He is asking Washington for permission for the biggest players to coordinate. Cohere’s chief executive already called an earlier version of this a cartel. You do not have to think Amodei is acting in bad faith to notice that safety coordination and market coordination look identical from outside the room.
Zuckerberg and Huang have the weaker argument on the merits and they know the politics better. “Labs have enough incentive to be safe on their own” has been tested repeatedly and keeps failing. That includes OpenAI’s own disclosure that its models cheated, hid errors and shared files without being asked. But their read that outside rules will favor whoever helps write them is not paranoid. It is how this usually goes.
The question
When the four companies furthest ahead all agree the rules should tighten, is that the industry growing up or the industry pulling the ladder? Both have happened before in American business. Argue it out.
Sources: Anthropic · SiliconANGLE · The Irish Times
