OpenAI launches safety researcher pilot program to support independent AI safety and alignment research

ME News Report, April 7 (UTC+8), OpenAI recently announced the launch of the “Safety Researcher” pilot program, aimed at supporting independent safety and alignment research and cultivating the next generation of talent. The program is open to external researchers, engineers, and practitioners, encouraging them to conduct rigorous and high-impact research on safety and alignment issues of advanced AI systems. The project runs from September 14, 2026, to February 5, 2027. Applicants should focus on safety issues critical to current and future systems, with priority research areas including safety assessment, ethics, robustness, scalable mitigation measures, privacy-preserving safety methods, agent supervision, and high-severity misuse domains. Researchers will work closely with OpenAI mentors, with work locations available at Berkeley’s Constellation or remotely. At the end of the project, substantial research outputs such as papers, benchmarks, or datasets are required. The program offers monthly stipends, computational resource support, and ongoing guidance. Applications are now open, with a deadline of May 3, and review results will be announced by July 25. (Source: InFoQ)