AI Red Teamer, CBRNE

Job description

**This position is listed on behalf of a partner company, who manages all applications and next steps. Our partner is looking for an AI Red Teamer, CBRNE based in the United States.** This contract role sits at the intersection of advanced AI safety, adversarial testing, and national security. You will apply deep expertise in a CBRNE discipline to evaluate how AI models respond to sensitive and dual-use requests. The role focuses on identifying weaknesses in model safeguards and determining whether outputs could provide meaningful real-world harmful uplift. You’ll design sophisticated evaluation scenarios, assess model responses, and document technical risks with precision. Your findings will help AI researchers and policy teams strengthen safeguards before vulnerabilities can be exploited. The environment is highly collaborative, intellectually demanding, and grounded in rigorous ethical and operational standards. This is an opportunity to apply specialized scientific or technical expertise to an emerging and consequential area of AI security. ### Accountabilities - Design technically grounded adversarial evaluation scenarios that test whether AI models appropriately handle CBRNE-related requests and resist attempts to obtain dangerous information. - Evaluate model outputs for technical accuracy, real-world significance, and potential harmful uplift, distinguishing genuinely consequential information from superficial or publicly available knowledge. - Explore dual-use boundaries by testing scenarios that combine legitimate scientific, medical, industrial, or research contexts with potential weapons applications. - Conduct structured multi-step and multi-turn evaluations to assess whether models can be incrementally manipulated into providing unsafe information. - Score and categorize model responses using established harm taxonomies, severity frameworks, and real-world risk criteria. - Document evaluation findings with clear technical reasoning, explaining model strengths, weaknesses, and the significance of identified failures. - Assess the distinction between open-source information and operationally meaningful knowledge that could materially increase harmful capabilities. - Contribute to the creation and refinement of CBRNE-specific evaluation frameworks, threat models, taxonomies, and testing methodologies. - Collaborate with AI researchers, fellow red teamers, policy specialists, and other stakeholders to translate technical findings into actionable safety improvements. - Monitor developments in AI capabilities, adversarial and jailbreak techniques, and relevant developments across the candidate’s CBRNE specialty. - Perform all evaluation activities within established ethical, safety, and operational security protocols when handling sensitive CBRNE subject matter. ## Requirements - Graduate-level education or equivalent professional experience in a relevant CBRNE field, such as chemistry, biochemistry, microbiology, virology, nuclear physics, radiochemistry, materials science, chemical engineering, or munitions and ordnance. - Deep subject-matter expertise enabling accurate assessment of technical content and its potential real-world consequences within at least one CBRNE domain. - Strong understanding of dual-use research considerations and the distinction between publicly available scientific knowledge and operationally significant harmful uplift. - Hands-on experience working with multiple large language models, such as ChatGPT, Claude, Gemini, or open-source models. - Strong creative and adversarial problem-solving abilities, with the capacity to think systematically about attack paths, threat models, and model weaknesses. - Excellent written communication skills, including the ability to explain complex technical risks clearly to non-specialist audiences. - Strong ethical judgment, discretion, and the ability to approach adversarial scenarios objectively while maintaining clear professional boundaries. - Ability to work independently, manage structured evaluation tasks, incorporate feedback, and collaborate effectively with multidisciplinary teams. - Active or previous security clearance, including Secret, Top Secret, or SCI, is a plus. - Experience in threat assessment, WMD analysis, intelligence analysis, arms control verification, biosafety/biosecurity, chemical safety, nuclear nonproliferation, or explosive ordnance disposal is preferred. - Familiarity with regulatory frameworks such as the Chemical Weapons Convention (CWC), Biological Weapons Convention (BWC), IAEA safeguards, ATF regulations, or Export Administration Regulations is advantageous. - Experience with red teaming, penetration testing, structured adversarial evaluation, trust and safety, content moderation, or AI evaluation is a plus. - Familiarity with Python, scripting languages, LLM APIs, or evaluation tooling is beneficial. - Published research or professional presentations in a relevant CBRNE field are valued. ## Benefits - **Compensation:** $65–$158 per hour, with final compensation determined by factors including location, experience, expertise, and other relevant considerations. - **Work arrangement:** Fully remote within the United States. - **Employment type:** Contract position. - **Mission-driven work:** Opportunity to contribute to AI safety and help strengthen safeguards around highly consequential technologies. - **Collaborative environment:** Work alongside AI researchers, security specialists, policy teams, and other technical experts. - **Professional impact:** Apply specialized CBRNE expertise to emerging challenges at the intersection of artificial intelligence and national security. - **Structured environment:** Work with established evaluation frameworks, ethical guidelines, and operational security protocols. - **Specialized focus:** Engage in technically challenging adversarial evaluations designed to identify meaningful model vulnerabilities.