Remote AI Safety Red-Teamer | Adversarial Testing

Vettedlance; United States; Remote; USD 28,000 - 50,000; Part time

Job summary

Vettedlance in the United States is seeking a Video and Sensor Data Annotator for Robotics to probe AI models for unsafe, biased or misleading behavior. This is an hourly or per-task contract, remote, flexible and part-time, with weekly payments in USD.

You will write adversarial prompts, record failures with reproduction steps, rate severity, and suggest safe responses. The role is an independent contractor with rolling applications as projects open.

Qualifications

• Background in security testing, journalism, law or social research.
• Fluency in a second language, especially one poorly served by AI.
• Good judgement and discretion.

Responsibilities

• Write adversarial prompts across safety, bias and misinformation topics
• Record failures with clear reproduction steps
• Rate the severity of each failure against a policy
• Suggest what a safe response would look like
• Creative, persistent and methodical
• Comfort working with sensitive and difficult material
• Good judgement and discretion

Job description

Vettedlance in the United States is seeking a Video and Sensor Data Annotator for Robotics to probe AI models for unsafe, biased or misleading behavior. This is an hourly or per-task contract, remote, flexible and part-time, with weekly payments in USD.

You will write adversarial prompts, record failures with reproduction steps, rate severity, and suggest safe responses. The role is an independent contractor with rolling applications as projects open.