openaienmodel: gpt-5-mini-2025-08-07
Announcing the OpenAI Safety Fellowship
Key Points
- 6-month pilot: Sep 14, 2026–Feb 5, 2027
- Applications open until May 3; decisions by July 25
- Stipend, compute, API credits, mentorship; no internal system access
Summary
OpenAI is launching the OpenAI Safety Fellowship, a 6-month pilot program to support independent, high-impact research on the safety and alignment of advanced AI systems. The fellowship runs from September 14, 2026 through February 5, 2027. Fellows receive a monthly stipend, compute support, API credits, and ongoing mentorship, with workspace available in Berkeley (Constellation) or the option to work remotely. Fellows will not have internal system access.
Key Points
- Focus: rigorous, technically strong, and community-relevant work in safety evaluation, ethics, robustness, scalable mitigations, privacy-preserving safety methods, agentic oversight, and high-severity misuse domains.
- Deliverable: fellows must produce a substantial research output (e.g., paper, benchmark, or dataset) by program end.
- Support: monthly stipend, compute support, API credits, mentorship, and cohort engagement; workspace available in Berkeley.
- Timeline: applications open now, close May 3, 2026; applicants notified by July 25, 2026; program runs Sep 14, 2026–Feb 5, 2027.
- Application notes: letters of reference are required; selection prioritizes research ability, technical judgment, and execution over formal credentials. See the application form for full eligibility and compensation details.
- Contact: openaifellows@constellation.org for questions about the application process.