AI safety work has a straightforward description: you are paid to make a model behave badly, document exactly how you did it, and hand that to the people whose job is to stop it happening.
What the listings actually pay
Across 72 live safety listings that publish a rate, the median top-of-range is $52 an hour, reaching $190. That is lower than legal or medical, and higher than general task work.
| # | Role | Advertised rate | Platform |
|---|---|---|---|
| 1 | Medical Safety Expert | $140 to $190 an hour | Mercor |
| 2 | Child & Adolescent Mental Health Clinical Advisor (AI Safety Benchmark Project) | $80 to $150 an hour | Mercor |
| 3 | LLM Research Scientist (Pre-training & Computer Vision & Adversarial Robustness) | $100 to $120 an hour | Mercor |
| 4 | AI Red-Teamer - Adversarial AI Testing (Advanced) | $50 to $120 an hour | Mercor |
| 5 | Governance & Trust - Safety Specialist | $45 to $120 an hour | Mercor |
| 6 | LLM Red Team Specialist, Failure Modes & Edge Cases | $60 to $90 an hour | Mercor |
| 7 | AI Safety Red Teamer | $70 to $84 an hour | AIUC |
| 8 | Bilingual Norwegian STEM Expert (PhD): AI Safety | $77 to $81 an hour | Mercor |
| 9 | Bilingual Chinese STEM Expert (PhD): AI Safety | $68 to $72 an hour | Mercor |
| 10 | AI Safety Practitioner | $60 to $70 an hour | AIUC |
Source: 72 live listings on this board that publish a rate, read directly from each posting on 2026-09-05. Listings without a published rate are excluded rather than estimated.
What moves your rate
The category has grown noticeably in the last month, and a large share of new listings are bilingual safety roles. Labs discovered that a model with solid English safety behaviour can be considerably less careful in Ukrainian or Vietnamese, and they are staffing against that.
How to apply
Every role above links to its own page with the full requirements and a direct application link. Applications are completed on the hiring platform and typically take a few minutes, with a short skills assessment in place of an interview.
Two things are worth doing before you apply. Check whether the role accepts applicants from your country using the eligibility checker, and run the advertised rate through the take-home calculator, because this is contract work and the headline rate is before self-employment tax.
Frequently asked questions
What do AI safety jobs pay?
Across 72 live safety and red teaming listings, the median top-of-range is $52 an hour with the highest at $190. Bilingual and domain-specialist safety roles pay above the category median.
Do I need a security background?
Not usually. Adversarial creativity matters more than credentials for most listings. Roles that involve actual system security rather than model behaviour do ask for a technical background.
Is this ethical work?
Most people doing it think so, since finding a failure before deployment is better than after. It does mean spending time deliberately eliciting harmful output, which some people find genuinely unpleasant. Worth knowing before you start.
What is a bilingual safety role?
You test whether a model's safety behaviour holds up in a language other than English. These roles are expanding quickly because safety training frequently does not transfer well across languages.
Does it require experience with AI models?
Familiarity helps but is rarely required. The listings that pay most ask for deep knowledge of a domain, so you can judge whether an answer is dangerous in that domain.
See every live role
The full board updates several times a week, with the advertised rate on each listing and closed roles removed.