Safety Hiring Lags as Anthropic Pushes to Pace the Frontier

Anthropic is calling to pace the frontier, with support from OpenAI and Microsoft, at the same time frontier voices warn about RSI and agent swarms. The labor signal cuts another way. Enterprises are building out applied AI, not massing around safety teams, and the roles that are actually open are tilting toward modeling and data. Across the 193 employers we track, 4,206 AI roles were open as of 2026-09-14, down 315 on the week. Our model puts new AI jobs created month-to-date at roughly 75,459. Within that, safety hiring remains small, with only 40 safety jobs across 19 employers in our live count. That mismatch creates pressure: a public case for slowing capability runs up against organizations racing to operationalize it. For workers, it is a two-track market: the applied side is hot, while oversight, evaluation, and ethics roles could grow more slowly and selectively. If Anthropic’s plan shapes norms, companies may add governance and assurance muscle, but today’s headcount tells candidates to target modeling, data, and deployment skills. Leaders who talk RSI risk without budgeting for safety hires are setting up talent bottlenecks later.