Researcher, Misalignment Monitoring Lead

ABOUT THE TEAM OpenAI's Safety Research organization works to ensure that our most capable AI models can be deployed safely and responsibly. As frontier models become increasingly capable and autonomous, developing reliable ways to monitor their behavior, identify emerging risks, and evaluate the effectiveness of safety systems becomes increasingly important. Our work focuses on advancing the research and evaluation methods needed to understand model behavior and ensure that monitoring systems remain effective as model capabilities evolve. We combine rigorous empirical research with practical safety interventions, working closely with research, engineering, and deployment teams to translate findings into improvements across OpenAI's models and products. ABOUT THE ROLE We are looking for a senior researcher to lead technical work on misalignment monitoring for frontier AI systems. You will develop methods to evaluate the effectiveness of AI monitors, understand their limitations, and improve our ability to detect and stop concerning model behavior reliably. This role combines research leadership with hands-on technical execution. You will set research priorities, design rigorous evaluations, investigate monitoring failures, and develop approaches for stress-testing and improving monitoring systems. A key focus will be understanding where monitors produce false positives, miss important behaviors, or become less effective as models grow more capable. You will work closely with researchers and engineers across OpenAI to turn monitoring research into practical improvements to our safety systems. We are looking for someone with strong research judgment, technical depth, and the ability to independently drive ambiguous problems from initial investigation through measurable results. IN THIS ROLE, YOU WILL: - Lead the research and technical direction for evaluating and improving the misalignment monitoring of frontier AI systems. - Design and execute rigorous evaluations that measure monitor effectiveness, identify failure modes, and assess the reliability of monitoring signals. - Develop methods to stress-test and red-team monitors, including identifying behaviors that evade detection or expose weaknesses in existing approaches. - Investigate false positives, missed detections, and other monitoring failures to improve the accuracy and usefulness of safety evaluations. - Build scalable evaluation frameworks and experimental tools that enable ongoing measurement of monitoring performance as model capabilities evolve. - Translate research findings into actionable improvements, working with partner teams to strengthen monitoring systems and inform broader safety decisions. - Establish research priorities and guide technical execution, balancing foundational research questions with practical deployment needs. YOU MIGHT THRIVE IN THIS ROLE IF YOU: - Have a strong background in machine learning research, particularly in model evaluations, AI safety, alignment, robustness, or related areas. - Have demonstrated the ability to lead technically challenging research projects, from identifying important questions through designing experiments and delivering results. - Bring strong experimental judgment and understand how to design evaluations that are rigorous, reproducible, and difficult to game. - Have hands-on experience building or evaluating machine learning systems, including working with large language models or other frontier AI systems. - Enjoy investigating complex failure modes and developing methods to distinguish meaningful safety signals from misleading measurements. - Are comfortable working on open-ended research problems where the right methodology or success criteria may not be obvious at the out