RemoteWork from anywhere with a stable connection.
Flexible, part-time (10–40 hrs/week)Choose your own hours.
Ongoing, project-based
Starts when a matching project opens
· Rolling applications
Role overview
Specialists probe AI models for dangerous capabilities and unsafe behavior, and write the test cases and rubrics used in safety evaluations.
Key responsibilities
Design adversarial prompts in your specialty: CBRN, cyber threat intelligence, OSINT, child safety.
Run jailbreak and prompt-injection tests and document successful attacks.
Grade model answers for harmful uplift against safety rubrics.
Arabic track: safety evaluation of Arabic model outputs.
Ideal qualifications
One or more of: chemical, biological, radiological, nuclear or explosives safety background (3+ years); cyber threat intelligence practice; multilingual OSINT or web-intelligence analysis; hands-on LLM red-teaming; child-safety research or trust and safety work.
Sound judgement with sensitive material.
Nice to have
Native Arabic.
Relevant certifications (for example GIAC, OSCP, CIH).
Published safety research.
More about the opportunity
Some CBRN projects require residence in the US, UK, Canada, Australia or New Zealand; this is stated per project. Work involves sensitive content, and wellbeing support is provided.
Compensation and contract terms
$60 – $90 / hour. Paid in USD, weekly. Rate set after a short paid assessment.
Hourly contract. You are engaged as an independent contractor.
Remote. Flexible, part-time (10–40 hrs/week), on your own schedule.
Paid weekly in USD for approved work.
Projects can be extended, shortened or ended early depending on project needs and quality.
Application process
ApplyCreate a free profile and upload your CV.
InterviewAbout 20 minutes. You need a laptop or desktop with a camera and microphone.
ID and credential checkWe verify your identity and the qualifications the role needs.
Matched to projectsOnce verified, you are invited to paid projects that fit your skills.
Vetting for this role: Profile and CV review, AI interview in English or Arabic, proctored (about 20 minutes), Identity check: government ID and liveness, Credential and employment verification, Short paid assessment and Expert Score.
Free to apply. Vettedlance never charges candidates, and will never ask you to pay for training, tests or equipment.
Related roles
Cybersecurity Practitioner Evaluator
Hiring now
$75 – $95 / hour
Defence & security
Expert · Open now
Military Operations Veteran, AI Scenario Evaluator
Hiring now
$40 – $80 / hour
Defence & security
Expert · Open now
Intelligence Analyst (HUMINT, SIGINT) for AI Evaluation
Hiring now
$50 – $100 / hour
Defence & security
Expert · Open now
Imagery Analyst, Drone & Maritime AI Training Data (US)