OpenAI is kicking off a pilot program called the OpenAI Safety Fellowship, designed to support external researchers and engineers working on high-impact safety and alignment research for advanced AI. The program runs from September 2026 to February 2027 and focuses on priority areas like safety evaluation, ethics, robustness, and scalable mitigations. Fellows will get mentorship from OpenAI experts, a monthly stipend, compute support, and workspace in Berkeley (with a remote option). They’re expected to produce a concrete output like a paper or benchmark by the end. OpenAI is looking for applicants from diverse fields including computer science, social science, cybersecurity, and HCI — valuing research chops over formal credentials. Applications close May 3, with notifications by July 25. Why it matters: This signals OpenAI’s recognition that AI safety is too big for any one lab to solve alone, and that nurturing independent talent is key to ensuring powerful systems stay aligned as they evolve.