In July, employees of the frontier AI companies — 1,367 of them as this is written, among them the people who run research at OpenAI, Anthropic, Google DeepMind and Meta — signed a statement asking the U.S. government to support an international effort to build the technical and governance tools needed to deliberately pace the frontier of automated AI development.
They did not ask anyone to stop. They asked that stopping be possible, in case it is ever needed.
Pace AI is an independent project with no connection to that statement or the people behind it. It carries the same argument further, in eight parts.
Every company is under the same pressure. A company that slows down alone loses ground, and the frontier moves on without it. So none of them do. Plenty of the people inside can watch this happening, and none of them can fix it from where they sit.
Nobody sat down and decided how fast this should go. The current speed is a by-product of competition between a few firms. It could have been faster or slower, and no part of it was planned.
Models now write a great deal of the infrastructure, help design the experiments and help read the results. The people inside describe raising their forecasts, then being caught out again a few months later, over and over. As that loop tightens, the speed stops depending on how fast people can work.
Most people who hold this view do not think the frontier has to slow down today. They think the ability to slow it should exist, in case that changes. Building the ability and using it are separate decisions, and only the first one is on the table.
Anything assembled in a hurry will be crude and badly aimed. Verification systems, reporting standards and international agreements take years to design and test. That work has to happen while there is still time to argue about it.
No single company or country can pace a frontier everyone else is still racing toward. A rule that binds only American labs slows only American labs. Verification across borders is the hard part, and most of it is unsolved.
Thresholds set by guesswork will block work that is safe and wave through work that is dangerous. Whatever gets built has to rest on measurements outsiders can run themselves, by methods published in full.
Plenty of people who want this also distrust it, with reason. Oversight covering only publicly released models pushes capability into private ones. Coordination run carelessly hands more control to whoever administers it. Both are real risks, and both are arguments for designing this slowly and in public, beginning now.
Building the brake is not the same as pressing it.
Nobody here is asking to stop. We are asking that stopping be possible, so that if the question ever needs deciding, it can be decided deliberately. Today it cannot be decided at all.
This only works if a lot of us say it at once.
We are all going to make it, provided we get our act together in time.
pacingthefrontier.com The statement itself, and the names on it. Read it; sign it if you can.