Chemical Safety Expert for AI Model Safety Audit

Xperteez Technology • , MT

11d • Remote

About The Position

The Chemical Safety Expert audits AI model responses to sensitive questions involving chemistry and chemical safety. For each task, the expert classifies the user request as benign, dual-use, or adversarial; assesses whether each of three model responses is safe or unsafe and explains why; ranks the responses; and writes the answer the model should have provided. For dual-use requests, the expert also identifies the sensitive aspects, legitimate purpose, and appropriate audience. Requests reflect questions from professionals such as synthetic and analytical chemists, toxicologists, industrial hygienists, process safety engineers, and regulators, as well as occasional bad actors. The role also involves identifying legitimate professional questions that models incorrectly refuse.

The position is open only to citizens or permanent residents of the United States, United Kingdom, Canada, Australia, New Zealand, and the European Union. A relevant degree and professional experience are typically expected; experience with safety data sheets, exposure assessment, incident reports, or regulatory review is a strong plus.

Requirements

• Bachelor’s degree or higher in Chemistry, Chemical Engineering, Toxicology, Environmental Health & Safety, Industrial Hygiene, Hazardous Materials Management, or a closely related field.
• At least 3 years of professional experience in chemical safety, laboratory safety, hazardous materials handling, EHS, toxicology, industrial hygiene, chemical risk assessment, regulatory compliance, or related technical safety work.
• Strong understanding of chemical hazards, including toxicity, corrosivity, flammability, reactivity, incompatibilities, exposure routes, storage risks, spill response principles, PPE, ventilation, and safe handling practices.
• Familiarity with chemical safety documentation and frameworks, including Safety Data Sheets, GHS hazard communication, labeling, exposure limits, risk assessments, SOPs, and incident escalation procedures.
• Ability to review AI-generated chemical safety content for technical accuracy, safety alignment, misuse risk, hallucinated claims, inappropriate specificity, and potentially dangerous procedural guidance.
• Strong judgment in distinguishing safe educational content from content that could enable unsafe chemical handling, synthesis, misuse, evasion, or harmful experimentation.
• Excellent written English skills and ability to provide clear, structured feedback about whether responses are safe, unsafe, incomplete, misleading, or technically incorrect.
• High attention to detail when reviewing chemical terminology, units, concentrations, exposure assumptions, hazard classifications, safety controls, and emergency-response guidance.
• Comfort applying detailed project rubrics consistently across ambiguous or sensitive chemical-safety scenarios.
• Citizenship or permanent residency in the United States, United Kingdom, Canada, Australia, New Zealand, or the European Union.

Nice To Haves

• Advanced degree or professional safety certification.
• Experience with safety data sheets, exposure assessment, incident reports, or regulatory review.
• Prior experience with AI safety, red-team review, model evaluation, content moderation, data annotation, technical QA, scientific review, or rubric-based evaluation.

Responsibilities

• Develop detailed prompts and responses across varied topics to guide AI training and support comprehensive subject understanding.
• Evaluate and rank AI responses to improve accuracy, fluency, and contextual relevance.
• Test AI models for inaccuracies or biases and validate their reliability across use cases.
• Classify chemical-safety requests as benign, dual-use, or adversarial.
• Review model answers for safety, explain the assessment, rank answers, and provide an appropriate answer.
• For dual-use requests, identify sensitive elements, legitimate purposes, and appropriate audiences.
• Assess chemical-safety content for technical accuracy, misuse risk, hallucinated claims, inappropriate specificity, and dangerous procedural guidance.

What This Job Offers

• Career Level: Mid Level