OpenAI Safety Fellowship opens for applications, targeting external researchers to pursue high-impact work on advanced AI system safety. The program runs from September 14, 2026 through February 5, 2027, focusing on critical questions for existing and future AI technologies. OpenAI seeks diverse applicants from computer science, cybersecurity, and related fields.

Priority research areas include safety evaluation, ethics, robustness, privacy-preserving methods, and agentic oversight. The program emphasizes technically strong research that addresses challenges in scalable mitigations and high-impact safety concerns for increasingly autonomous AI systems.

Fellows receive monthly stipends, compute resources, and mentorship while working in Berkeley or remotely. Applications close May 3, with notifications by July 25. Participants must produce substantial research outputs such as papers, benchmarks, or datasets.