A six-month fellowship funding external researchers, engineers and practitioners to work on AI safety problems including safety evaluation, robustness, privacy-preserving safety methods, agentic oversight, ethics and high-severity misuse domains. Fellows receive a stipend, access to OpenAI models and API credits, compute support and ongoing mentorship, and are expected to produce a substantial research output such as a paper, benchmark or dataset. Workspace is available at Constellation in Berkeley, California, with remote participation possible. International candidates may apply, and OpenAI states it weighs research ability, technical judgement and execution over specific credentials.