OpenAI
Researcher, Agent Safety, Oversight and System Mitigations
$380K–$500K
San Francisco · Full-time · Posted Sep 3, 2026
About the job
Designs, builds, and evaluates oversight systems and system-level controls to help AI agents act safely while preserving useful autonomy.
What you'll do
- Build and evaluate controls for agent actions, including their integration with sandboxing, process isolation, and permission boundaries.
- Work with Codex harness engineering to productionize AI controls and red-team end-to-end agent systems.
- Measure and improve the safety–productivity tradeoff, including missed harmful actions, unnecessary blocks, approval burden, and latency.
What they ask for
- Strong systems or security instincts, including concrete reasoning about isolation boundaries, permissions, attack surfaces, and failure modes.
- Ability to develop threat models, reproducible experiments, and practical mitigations.
- Ability to build experimental infrastructure and design evaluations.
Nice to have
- Background in AI control or security is welcome but not required.
- Deep interest in frontier AI alignment, safety, and control.
Pay
$380K – $500K USD base salary
Tools
Codex
AI focus
Developing and evaluating AI agent oversight, action controls, and system-level safety mitigations.
Model-written summary · the original posting is authoritative