Policy and Toxicity Evaluator

Overview We are seeking Policy and Toxicity Evaluators to support a pilot project focused on analyzing and grading AI-generated model outputs for a leading AI platform. In this role, you will assess model responses to ensure compliance, safety, and adherence to project-specific guidelines. Evaluators will work across three workflows: refusal analysis, toxicity detection, and policy verification, helping to validate the model’s behavior and improve its safety and utility.  What you will do:  • Analyze AI model outputs to assess quality and safety.  • Identify instances where the model refuses to answer a prompt and determine if the refusal was necessary or an error (over-refusal). • Identify toxic content such as hate speech, harassment, explicit material, or self-harm encouragement. • Review model responses for compliance with project-specific policy guidelines. • Shift between refusal, toxicity, and policy workflows as needed based on project volume. Requirements:  • Excellent command of written English to process complex prompts and long-form model responses quickly.  • Strong critical thinking and the ability to make objective decisions.  • Exceptional attention to detail, including identifying borderline violations.  • Comfort navigating ambiguous or complex content requiring deep judgment.  • Ability to internalize and strictly follow complex policy guidelines.  • Prior experience in Trust & Safety, Content Moderation, or RLHF (Reinforcement Learning from Human Feedback) annotation is highly beneficial.    Project Details:  Contract Type: Freelance  Location: Philippines (remote) Duration: 1 to 2 weeks Schedule: 10 hours weekly; flexible based on client's needs   Note: Please do not use VPNs or IP-masking tools during the recruitment process — our security system requires accurate regional verification.