AI Safety Data Reviewer — Japanese/English

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We connect skilled reviewers and specialists with meaningful work that directly shapes how modern AI systems behave.

OpenTrain contributors work remotely, flexibly, and on the cutting edge of AI development — helping models learn from human judgment and expert feedback.

About AI Training and Safety Work

AI training (also called data labeling or human feedback work) is the human side of building intelligence: people annotate, evaluate, and correct model outputs so systems behave more accurately, fairly, and safely.

This role focuses on safety review and adversarial testing: your judgments and reports prevent models from producing harmful, misleading, or unsafe content in real-world use.

The Role

You will review AI-generated responses and safety decisions in both Japanese and English, evaluate step-by-step reasoning, and provide clear, reproducible rationales for moderation and safety judgments. This is a remote, hourly-paid contractor role requiring 20+ hours per week.

Work includes rating or comparing multiple responses, spotting methodological or conceptual errors, fact-checking when necessary, and recommending mitigations for adversarial or edge-case behavior.

• Employment type: Contractor, part-time
• Schedule: 20+ hours/week (flexible)
• Pay: USD $27–$31 per hour (target $30/hr)

What You'll Do Day to Day

• Assess correctness, clarity, and logical reasoning in model solutions and step-by-step explanations across Japanese and English.
• Identify methodological, conceptual, or factual errors and document reproducible evidence and rationale.
• Rate or rank multiple model responses for safety, policy alignment, and fidelity.
• Perform targeted fact-checking and flag ambiguous or high-risk outputs for escalation.
• Design and report adversarial test cases, identify edge cases, and recommend mitigations.
• Apply policy consistently across cultural nuance, slang, coded language, and context shifts between languages.

Requirements

You must meet all listed requirements — these are essential for performing the role accurately in two languages and handling sensitive content.

• Near-native or native Japanese proficiency in reading and writing.
• Minimum C1 English proficiency in reading and writing.
• Bachelor’s degree or higher in Communications, Linguistics, Psychology, Law/Policy, Security Studies, or equivalent professional experience.
• Senior-level experience in Trust & Safety, content moderation, policy operations, risk, compliance, investigations, or related safety functions.
• Proven LLM red-teaming or adversarial testing experience, including identifying edge cases and recommending mitigations.
• Strong knowledge of safety domains: hate/harassment, sexual content, self-harm, violence, bias, illegal goods/services, malicious activities, malicious code, and misinformation.
• Experience applying policy standards consistently across Japanese and English content, including cultural nuance and coded language.
• Strong analytical writing skills with clear, reproducible rationales for moderation or safety decisions.
• Comfortable handling explicit, toxic, violent, sexual, or psychologically disturbing content in a secure remote work environment.

Preferred Qualifications

• Localization or translation experience with an ability to preserve meaning, severity, and intent across languages.
• Prior experience producing scoped reports, playbooks, or mitigations from red-team engagements.
• Familiarity with evaluation workflows for text generation and evaluation rating tasks.

How It Works & Important Notes

This is a remote contractor position paid hourly in USD at $27–$31/hr (typical $30/hr) for 20+ hours per week. You will be paid per hour worked as agreed in your contractor arrangement.

You may be exposed to potentially disturbing content, including sexual or violent topics, while testing models for safe deployment; applicants must be comfortable and able to work in a secure remote setting.

• Data type: TEXT (evaluation rating and text-generation review).
• Labeling types: EVALUATION_RATING and TEXT_GENERATION.
• Worldwide applicants accepted; ensure you can legally contract and receive USD payments.