Researcher, Recursive Self-Improvement Safety

ABOUT THE TEAM Models 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 OpenAI’s ability to continue deploying models safely. The Preparedness team is dedicated to addressing these critical risks. Our work includes: 1. Measurement. Monitoring and predicting the evolving capabilities of frontier AI systems. 2. Mitigation. Keeping misal… Full description ↗