Head of Model Safety at a frontier model lab in the middle of its second flagship training run.

A frontier model lab in the middle of its second flagship training run. Funding is comparable to the largest UK and US labs. Around 180 staff, of which roughly 100 are research or applied research. The lab publishes on alignment and safety as part of its public commitment; the Head of Model Safety is a named voice in that publishing.

We name a client on the mandate page only where the firm has approved publication. Where not named, the description above is intended to be enough for a senior candidate to recognise whether their pattern fits the brief.

Head of Model Safety, reporting to the CEO. The seat is a newly created role lifted out of the Head of Research's remit as the safety programme has scaled.

Direct line management of around 18 safety researchers and engineers across three squads — capability evaluations, alignment research, and red-teaming infrastructure. Hiring four further senior researchers in the first nine months.

The 12-month remit is to define the lab's safety publication agenda, hold the line on pre-deployment evaluations for the next flagship model, and own the lab's engagement with the UK AI Safety Institute and equivalent US bodies.

Location: London or San Francisco. Travel between the two and to relevant policy convenings expected.

The Head of Model Safety is the lab's public voice on alignment and safety, with the post-holder named as the byline on the lab's published research in this area. Reporting is to the CEO. The work spans both the external publishing and the internal safety authority that sits alongside the lab's research function. Both halves are carried by one person rather than split across two roles.

— Peter Wood, Partner & Chief Strategy Officer