About the TeamOur Cyber team builds AI systems and products that help trusted defenders understand and respond to cyber threats while improving the safety and reliability of frontier models in security-sensitive settings.
The team works across product engineering, model training, evaluations, safeguards, and deployment to make advanced cyber capabilities useful to defenders and responsibly managed. We collaborate closely with Safety/Preparedness, Research, Security, Legal, Communications, GTM, and external partners across OpenAI's broader cyber work.
About the RoleWe're looking for research and software engineers to join Codex Cyber. You'll help define and ship security products, work with trusted defenders and customers, shape model training and access patterns, and build research and evaluation systems for assessing cyber capabilities, validating safeguards, and improving training data. This role is hands-on and cross-functional, connecting product launches, model development, safety work, and real-world security use cases.
In this role, you will:- Help define and execute the technical roadmap for Codex Cyber's security products, including evaluations, safeguards, trusted-defender workflows, and deployment decisions.
- Work with trusted defenders, customers, and partner teams to understand cyber use cases, evaluate risk, and turn feedback into product and research priorities.
- Shape cyber-specific model training and access patterns, including data, evaluations, validation, and deployment criteria.
- Build and validate systems for measuring cyber capabilities, monitoring misuse risk, and proving safeguards work in practice.
- Collaborate with Safety/Preparedness, Research, Security, Legal, Communications, Go-to-Market, and external partners on company-wide cyber priorities.
- Translate frontier cyber research into launch-ready tools, operational playbooks, and durable infrastructure for Codex and security products.
You might thrive in this role if you:- Enjoy 0 -> 1 environments, can navigate ambiguity, are excited to build security products that bring frontier AI capabilities to trusted defenders responsibly.
- Can move fluidly between product launches, customer and defender partnerships, model training, evaluation science, and security research.
- Have strong programming skills in Python, TypeScript/JavaScript, or similar languages, with the judgment to build reliable systems in ambiguous, high-stakes domains.
- Think clearly about cyber capability, misuse risk, access patterns, safeguards, and deployment tradeoffs in high-stakes environments.
- Have experience in ML systems, security products, cyber evaluations, model training, safety engineering, or trusted customer deployments.
