​🇺🇸 OpenAI Safety Fellowship: a paid fellowship in AI safety, BerkeleyA program from OpenAI for those who want to transition into full-time empirical AI safety research. Fellows will spend around 5 months working on their own research alongside mentors from OpenAI's safety and alignment teams. The final deliverable is a paper, benchmark, or dataset. Fellows work with external infrastructure: open-source models and public APIs. Remote participation possible, but in-person is preferred. 🙋‍♀️Who it's for:• people looking to transition into full-time empirical AI safety research;• with strong technical fundamentals: confident in Python, able to break ambiguous problems into concrete steps;• who thrive in fast-paced, collaborative environments and can drive projects forward under uncertainty;• with backgrounds in physics, math, CS, cybersecurity, or other quantitative fields;• a PhD, prior ML experience, or published papers are NOT required.📌 Duration: September 14, 2026 — February 5, 2027 (full-time, 40 hrs/week)💸 Stipend: $3,850/week + up to $15,000/month in compute. OpenAI supports J-1, F-1/CPT, OPT visas (other situations considered on request)Deadline: May 3, 2026 (11:59 PM, Anywhere on Earth)Good luck!