Three men who fight over every dollar, every benchmark, and every headline spent a single weekend agreeing to stop racing — and the smartest people in the room did not hear a safety breakthrough. They heard a bell. Chamath Palihapitiya, the billionaire who made his fortune early on the very companies now wielding the slowdown, called the “pace the frontier” pact what it actually is: an argument to end open source and concentrate the future in a handful of already-dominant hands. A voluntary slowdown with no enforcement, no published rules, and no seat at the table for the people it will actually hurt is not a safety plan. It is a cartel with a warning label.
The pact itself is real. In the span of roughly forty-eight hours, Anthropic chief Dario Amodei published his “We Must Pace the Frontier” essay, and the leaders of OpenAI, xAI, and Google DeepMind fell in line behind him. Sam Altman agreed, Elon Musk agreed, Demis Hassabis agreed. Four rival executives coordinating on anything is remarkable; that the one thing they coordinate on is slowing their own race should have set off every alarm in the building.
Chamath Palihapitiya — The Pact Is a Moat Wearing a Warning Label
Palihapitiya did not hedge. Writing on X, he charged that Amodei was really arguing for an end to open source and for concentrating “enormous technological and economic power over humanity” in the hands of the very firms signing the pledge. He is not wrong about who signs and who does not. Anthropic, OpenAI, and Google DeepMind all build closed models. The slowdown rules the trio are now championing would land hardest on the open-source camp — the startups, the small labs, and the independent researchers who do not have a compliance department, a government-affairs team, or a lobbyist on retainer.
A safety rule that only the companies with the biggest moats can afford to obey is not a rule. It is a fence, and they are building it around themselves.
Jason Calacanis — The Money That Cheered the Boom Is Now Running for the Door
The market heard the same bell. Jason Calacanis, the early-stage investor and podcast host, told his followers to brace for AI stocks to plunge more than ten percent on the Monday open, and he pinned the blame on a single blog post — charging that Amodei had upended the AI trade with one essay. Markets rattled in response. Not everyone bought the panic: Dan Ives, the tech analyst, waved it off as a blip, arguing the five trillion dollars of expected AI spending keeps the industry’s momentum intact no matter what four CEOs said on a weekend.
The same venture capital that begged founders to move faster is now measuring the distance to the exits — which tells you the pact frightened the one constituency that never admits anything is frightening.
The Open-Source Builder — “Slow Down” Is Easy to Say When You Already Hold the Lead
The real cost does not show up in a stock ticker. It shows up in a maintainer’s inbox. The open-source developer who has spent years building a model or a tool the giants now want slowed gets no chair at the table where four billionaires decide the speed limit. They get a new rule they did not vote on and a compliance bill they cannot pay. When the largest labs — the ones whose names are on the pledge — ask the world to hit the brakes, the brakes are attached to everyone else’s car.
The deal is also, by the signatories’ own admission, loose rather than legally binding, with each leader signing on only “at least partially.” A safety pact with no enforcement mechanism and no published terms is not a treaty. It is a gentlemen’s agreement — and in this particular room, the gentlemen already own the market.
The skepticism is not coming from cranks. It is coming from inside the tent. Former White House AI adviser David Sacks has argued that the guardrails the frontier labs describe look more like a competitive moat than a safety measure, and that coordination among a handful of dominant firms over who gets access deserves the same antitrust skepticism regulators apply to any other form of market allocation. He is describing the difference between a safety standard and a cartel in plain terms: published rules, a route to appeal a rejection, and a vote for the people the rules will hurt.
The story of the weekend, in the end, is not whether the machine is dangerous. The machine is dangerous; the people building it have been telling us so for a year. The story is who gets to decide what happens next — and whether a pact negotiated over a weekend by four men who already hold the lead becomes law for everyone still trying to catch them. Right now the answer to all three questions is no. The open-source developer does not get a chair, a vote, or a warning. They get a speed limit they never agreed to, set by competitors who were already ahead of them when the race was called.
There is a version of this story where the weekend was a genuine act of responsibility, four rivals finally admitting the thing they built is outpacing their ability to control it. That version deserves to be taken seriously, because the alternative — that the safety talk is a shield for market power — is a far uglier thing to believe about the people who own the future. But belief is not the point. Proof is. And the proof of whether this pact is safety or cartel will be written in the fine print that has not been published yet: who is admitted, who is refused, and whether the man denied a license gets to ask why. Until that paper exists, the only people slowed down by the frontier’s new speed limit are the ones who were already losing the race.
Sources: Semafor, Seoul Economic Daily, SiliconANGLE, AIToolly.