Safety roles
55 open safety roles from 12 networks. 29 of them (53%) say what they pay, most often $40+/hr - the rest disclose nothing until you apply.
Adversarial work: finding the prompts that get a model to do what it shouldn't, then labelling why the failure happened. Includes bias and policy review, and the judgement calls about where acceptable behaviour ends.
Frequently asked questions
What is Safety AI-training work?
Adversarial work: finding the prompts that get a model to do what it shouldn't, then labelling why the failure happened. Includes bias and policy review, and the judgement calls about where acceptable behaviour ends.
How many Safety roles are currently open?
55 right now, refreshed daily from the source listings, spread across 12 networks at roughly 5 each. Counts move daily as listings expire, so treat this as a snapshot rather than a trend.
Which companies hire for Safety roles?
12 networks are hiring: AI Community jobs board, Alignerr, DATAmundi freelance & vendor applications, DataForce jobs, Handshake AI, Mercor Experts, and 6 more. Each links to its own profile, where the ownership and pay claims carry their sources.
What do Safety roles pay?
Where pay is disclosed, the most common range is $40+/hr (20 of 55 roles). 26 of 55 listings don't publicly disclose pay.
Are Safety roles remote?
Location varies by role - the most common listing type is "Region specific" (27 of 55 roles).