The AI builders want a usable brake

Subject: The AI builders want a usable brake

Preview: More than 1,100 workers are not asking for a blanket stop. They want the option to buy time together.

More than 1,100 employees from frontier AI labs and technology companies signed the Pacing the Frontier statement. They are not asking to halt all AI work. They want technical and governance mechanisms that would let countries and companies slow down together if automated AI research begins to accelerate too quickly.

The statement asks the US government to support an international effort. It starts from a practical dilemma: one company may believe slowing down is sensible while fearing that a competitor or another country will keep accelerating.

The warning comes from people building the systems. That turns an abstract safety debate into an operating question: which capability should trigger a slowdown, who decides, and how can everyone verify that the same rule is being followed?

The signatories speak personally; their employers are not automatically committed. The statement does not yet create an enforceable mechanism. Its importance is the coordination problem it makes explicit, not a policy that has already been adopted.

Define your organisation's own slowdown thresholds: loss of control, excessive access, irreversible actions or untested capabilities. Decide in advance who can stop a system, what evidence is required and how restart approval works.

A brake matters only when you know who can use it, which signals activate it and what evidence is required.