The Signal · Column 065

OpenAI wants a rulebook, and that changes who you hire

Andrei van Noordt · 21 September 2026 · 1 min read

OpenAI is asking Washington to lead a global AI standards push, and the request tells you more about the next hiring cycle than about politics. A company routinely described as moving faster than its regulators is now asking for a rulebook. Not fewer rules. One set of them, written in a room it can stand in, rather than thirty sets arriving out of sequence from thirty capitals.

The reading most people miss is that a standard is not a brake. It is a specification. The moment one exists, a long list of open questions about evaluation, documentation, provenance and safety testing stops being a matter of taste and becomes a matter of conformance. Companies that treat that as a legal problem will hand it to counsel and wait. Companies that treat it as an engineering problem will find it needs people who can build to a spec that has not been published yet.

A standard does not slow the work down. It moves the work from opinion to evidence, and evidence has to be built by someone.

That is where the market reprices. For the last few years the scarce AI hire was the person who could get a model into production. The scarce hire of the next few is the person who can prove what that model does, in writing, to someone outside the company. Evaluation engineers, model risk specialists, the rare technologist who reads policy drafts without flinching. They sit between two departments that have never shared a budget line, which is exactly why almost nobody has been hiring them and exactly why they take months to find.

For hiring leaders, the practical read is simple. Start that search before the rule text is final, not after. The morning a standard is published, every company in your sector discovers the same gap at the same time, and the few people who can close it will already be halfway through three other processes. Build the proof layer while it is still quiet enough to hire into.