Inside the talent war for AI safety officers, a job that barely existed two years ago

SAN FRANCISCO — Two years ago, "AI safety officer" was a punch line or a conference badge. This autumn it is a line in the budget. Large U.S. companies — not only model labs — are hiring officials whose job is to slow a product, document a risk, and take the call when a customer or a regulator asks who is in charge. Compensation, recruiters say, now rivals product leadership in a handful of searches.

The demand is coming from three directions. Regulators want a named human. Insurers want a control function they can underwrite. Enterprise procurement teams, especially in health, finance, and the federal contractor universe, want someone to sit across the table who does not report into the growth number.

That last point is the cultural fight. Safety roles that report into the CEO or the general counsel behave differently from those that report into the CMO. Several candidates described walking away from searches where the mandate was "enable the launch." They are looking for the authority to delay it.

Whether the title becomes a profession or a fad will be decided by the first serious incident that a company survives — or does not. Until then, the labor market is doing what it always does at the birth of a function: overpaying for scarce people and hoping the org chart catches up.