claudeenmodel: claude-sonnet-4-20250514
OpenAI Launches Safety Fellowship Program for External Researchers
Key Points
- New fellowship program for AI safety research launching September 2026
- Monthly stipend and compute resources provided to external researchers
- Applications due May 3 with focus on empirically grounded safety work
Summary
OpenAI has announced the OpenAI Safety Fellowship, a pilot program running from September 14, 2026 through February 5, 2027, designed to support independent researchers in AI safety and alignment work.
Key Points
- Research Focus: Priority areas include safety evaluation, ethics, robustness, scalable mitigations, privacy-preserving safety methods, agentic oversight, and high-severity misuse domains
- Program Structure: Fellows work with OpenAI mentors, engage with peer cohorts, and are expected to produce substantial research outputs (papers, benchmarks, or datasets)
- Resources Provided: Monthly stipend, compute support, API credits, ongoing mentorship, and workspace at Constellation in Berkeley (remote work also available)
- Eligibility: Open to applicants from diverse backgrounds including computer science, social science, cybersecurity, privacy, and HCI
- Application Timeline: Applications open now, close May 3, 2026, with notifications by July 25, 2026
- Requirements: Letters of reference required; emphasis on research ability and technical judgment over credentials