OpenAI Safety Fellowship Program Announcement

OpenAI has introduced the OpenAI Safety Fellowship, a program designed for external researchers, engineers, and practitioners to conduct rigorous research on the safety and alignment of advanced AI systems. This initiative aims to integrate external technical expertise into the AI safety ecosystem to address critical risks associated with current and future AI models.

Program Timeline and Application Process

The OpenAI Safety Fellowship operates on a specific timeline for the 2026-2027 cycle. Applications open immediately and close on May 3, with successful applicants being notified by July 25. The fellowship program itself runs from September 14, 2026, through February 5, 2027.

Priority Research Areas

The fellowship prioritizes research that is empirically grounded and technically strong. OpenAI has identified several key domains of interest for applicants:

• Safety Evaluation and Robustness: Developing methods to test and ensure the stability of AI systems.
• Alignment and Mitigations: Researching scalable mitigations and alignment techniques.
• Oversight and Privacy: Focusing on agentic oversight and privacy-preserving safety methods.
• Ethics and Misuse: Addressing ethics and high-severity misuse domains.

Fellow Support and Deliverables

Fellows are expected to produce a substantial research output by the conclusion of the program, such as a dataset, a benchmark, or a research paper. To support this work, OpenAI provides the following resources:

• Financial and Compute Support: A monthly stipend and compute support, including API credits and other appropriate resources.
• Mentorship: Close collaboration with OpenAI mentors and engagement with a peer cohort.
• Workspace: Physical workspace is available in Berkeley at Constellation, though remote work is permitted.

Eligibility and Access Constraints

OpenAI is seeking applicants from diverse academic and professional backgrounds, including cybersecurity, HCI, social science, computer science, and privacy. The selection process prioritizes technical judgment, execution, and research ability over specific credentials.

Regarding system access, OpenAI specifies that while fellows receive API credits and resources, they will not be granted internal system access.