jobs

Researcher, Recursive Self-Improvement Safety

About the teamModels are becoming increasingly capable—moving from tools that assist humans to agents that can plan, execute, and adapt in the real world. Mitigating the frontier risks resulting from these capabilities is paramount to OpenA

Company

OpenAI

Location

San Francisco; London, UK

Status

Open

Posted

2026-08-07T20:45:25.654+00:00

OpenAI's Preparedness team is hiring a researcher to prepare for risks from accelerated AI development, including recursive self-improvement. The work spans scalable oversight, automated auditing, red-teaming of misbehavior monitors, model behavior experiments and evaluations, and mechanisms for verifying compliance with future safety agreements. The posting wants strong technical executors with strategic and research taste who can turn open-ended objectives into concrete projects, build prototypes quickly, win buy-in across the company, and turn findings into production safety pipelines. Prior work in ML research, alignment or verification is a bonus.