AI Safety Red Team and Policy Evaluator

You will help evaluate and improve safety behavior in large language model systems. The work combines red teaming, policy-based content review, classification, and written evaluation.

You will review single-turn image edit requests and their outputs against a detailed safety taxonomy. You will also document safeguard bypasses, subtle policy violations, unclear guidance, and conflicts in the rules.

• Create adversarial test prompts that probe the limits of AI safety filters.
• Assess image edit requests and outputs for self-harm, violence, hate speech, sexual content, minors, identity depiction, and other safety categories.
• Apply safety guidelines consistently to ambiguous and borderline cases.
• Write self-contained evaluation rubrics and concise rationales that distinguish between model outputs.
• Optionally compare and rank multiple model outputs for the same request.
• Suggest clarifications when safety guidelines have gaps, ambiguities, or conflicts.

What It Pays and Takes

This is an entry-level, part-time contractor role. Relevant experience is strongly preferred, but the listing does not require a specific degree or prior job title.

• Pay: Not specified in the listing.
• Hours: 20 or more hours per week.
• Location: Open worldwide and fully remote.
• Language: Fluent in English.
• Equipment: Your own desktop or laptop and a reliable internet connection.
• Judgment: Ability to assess nuanced and ambiguous information against defined policy criteria.
• Red teaming: Experience designing challenge prompts, prompt engineering, or testing AI safety filters is helpful.
• Policy knowledge: Solid understanding of Trust and Safety principles for large language models.
• Writing: Ability to explain complex decisions clearly and concisely and create precise evaluation rubrics.
• Helpful background: A degree or equivalent experience in policy, law, ethics, linguistics, journalism, computer science, or a related field.
• Preferred experience: Content moderation, policy analysis, AI safety evaluation, RLHF workflows, or data annotation.

How It Works

Apply on OpenTrain with your resume, then complete the application on the hiring site. OpenTrain is the hiring and contracting organization for this role.

About AI Training Work

AI training is the human work behind systems that learn from examples, including reviewing model responses, testing safety behavior, and preparing labeled data. People with strong judgment and subject knowledge help turn difficult cases into clear training signals that improve how AI systems behave.