OpenAI
Researcher, Recursive Self-Improvement Safety
$380K–$500K
San Francisco +1 · Full-time · Posted Aug 7, 2026
- Location
- San Francisco
London, UK
About the job
Conduct research and build safety interventions and control systems to mitigate future misalignment and loss-of-control risks from recursive self-improvement.
What you'll do
- Develop and test risk assessments, monitoring, oversight, auditing, and other control measures for advanced AI systems.
- Turn open-ended safety goals into prioritized research, prototypes, and components of safety pipelines.
- Communicate findings, secure staff buy-in, and coordinate or manage others as needed.
What they ask for
- Exceptional technical execution.
- Strong strategic and research judgment, including prioritization in domains with weak feedback loops.
- Passion for mitigating risks associated with recursive self-improvement.
Nice to have
- Prior work in ML research, AI alignment, AI verification, or a listed domain.
Pay
$380K – $500K (base salary)
AI focus
Research and develop methods for scalable oversight, automated auditing, monitorability, model behavior evaluation, safety interventions, verification, and measurement of AI R&D risks related to future misalignment and recursive self-improvement.
Model-written summary · the original posting is authoritative