Overview
- OpenAI is recruiting for a new Safety Fellowship that backs independent work on how to test, control, and reduce risks in advanced AI systems.
- Selected fellows receive a $3,850 weekly stipend, about $15,000 per month in compute credits, and mentorship from OpenAI researchers, with no access to the company’s internal systems.
- Applications close May 3, notifications are due July 25, and the program runs from September 14, 2026 to February 5, 2027 with options to work from a Berkeley workspace or remotely.
- Each fellow must produce a paper, benchmark, or dataset in areas such as safety evaluation, robustness, scalable mitigations, privacy-preserving methods, agent oversight, ethics, or high-severity misuse.
- The launch came within hours of a New Yorker investigation that reported OpenAI had dissolved its superalignment, AGI Readiness, and Mission Alignment teams since 2024, shaping coverage of how the company approaches safety.