OpenAI Launches Safety Fellowship

OpenAI has announced a new initiative to cultivate external expertise in advanced AI safety and alignment. The OpenAI Safety Fellowship is a pilot program designed to support independent research in these critical areas.

Running from September 14, 2026, to February 5, 2027, the fellowship invites applications from researchers, engineers, and practitioners. The program aims to foster talent focused on safety questions relevant to both current and future AI systems.

Key Research Areas

Priority will be given to research proposals addressing safety evaluation, ethics, robustness, scalable mitigation strategies, privacy-preserving methods, agentic oversight, and high-severity misuse domains. OpenAI is particularly interested in work that is empirically grounded, technically rigorous, and beneficial to the wider research community.

Selected fellows will collaborate with OpenAI mentors and a cohort of peers. While workspace will be available in Berkeley, remote participation is also an option.

Program Benefits and Requirements

Fellows will receive a monthly stipend, compute resources, and ongoing mentorship. The expectation is for participants to produce a significant research output, such as a paper, benchmark, or dataset, by the program's conclusion. API credits and other resources will be provided, though internal system access will not be granted.

OpenAI encourages applications from diverse backgrounds, including computer science, social sciences, cybersecurity, privacy, and HCI. Emphasis is placed on research ability, technical judgment, and execution over specific academic credentials. Letters of reference are required.

Applications are open until May 3rd, with successful candidates notified by July 25th. For inquiries, applicants can contact [email protected].