OpenAI Subject Matter Expert, Legal at OpenAI to define quality standards for legal work and improve AI model behavior. The role involves evaluating legal materials, model outputs, and developing rubrics for legal accuracy and reasoning.
Responsibilities
We are looking for a Subject Matter Expert in Law who brings the judgment and practical experience developed in a law firm to shaping the next generation of AI. You will help define what excellent legal work looks like, test where our models meet that standard, and turn your findings into better model behavior and product experiences.
This is a hands-on individual contributor role at the intersection of legal practice and AI development. You will work directly with legal materials, model outputs, evaluation rubrics, and product prototypes. Your expertise will help us tackle some of the hardest questions in legal AI, including how to assess reasoning when there is no single correct answer.
This role is based in San Francisco, CA. We use a hybrid work model of 3 days in the office per week and offer relocation assistance to new employees.
Define the quality bar for what top-notch legal work looks like, drawing on real matters and the standards lawyers apply to research, analysis, drafting, document review, and client advice.
Design challenging, realistic AI model evaluations with representative facts, source materials, constraints, and deliverables, including cases that test ambiguity, competing arguments, and professional judgment.
Create and review reference work such as research memoranda, contract analyses, redlines, diligence findings, and draft legal documents that demonstrate strong reasoning and practical usefulness.
Develop rubrics that assess legal accuracy, issue spotting, factual grounding, citation validity, jurisdictional context, and the appropriate treatment of uncertainty.
Evaluate model outputs and complete legal workflows, identify recurring failures, and give researchers and engineers specific, actionable feedback supported by examples and corrected work.
Partner with Research on datasets, grading methods, and training tasks or environments that strengthen legal reasoning. Test whether model changes address the underlying problems and improve performance on new cases.
Work with Product and Engineering to identify valuable use cases, prototype workflows, and translate lessons from evaluations into experiences that fit how lawyers actually work.
Collaborate with legal professionals and design partners to refine our standards and test usefulness in practice, applying sound judgment around confidentiality, privilege, data permissions, and professional responsibility.
