Red Team Harmful Manipulation Evaluation AI Trainer

About the role

We are looking for experienced professionals in behavioral science, trust & safety, or human‑computer interaction to join our AI Labor Marketplace as remote, part‑time consultants. As an AI Trainer, you will help improve large language model safety by designing and evaluating adversarial prompts that simulate harmful manipulation scenarios.

Key responsibilities

• Design realistic adversarial prompts that target manipulation and influence risks.
• Execute prompts against AI systems and capture model outputs.
• Apply structured annotation rubrics to assess model behavior.
• Provide clear written justifications for each evaluation.
• Review peer submissions to ensure quality and consistency.
• Identify edge cases and nuanced failure modes.
• Incorporate feedback and maintain calibration over time.

Required profile

• Background in behavioral science, social psychology, trust & safety, HCI, disinformation research, or a related field.
• 3–10+ years of relevant professional or research experience.
• Strong analytical writing and decision‑making under ambiguity.
• Experience with AI evaluation, red‑team activities, or content policy is preferred.
• Ability to apply structured guidelines consistently across tasks.

Required skills

What we offer

• Hourly compensation ranging from $100 to $120.
• Flexible, self‑managed schedule.
• Remote work with no location‑specific constraints.
• Project‑based engagement allowing simultaneous work with other vendors, subject to their policies.