OpenAI has launched a new safety-focused fellowship, offering participants around $15,000 in monthly compute along with a $3,850 weekly stipend to support research on advanced AI systems.
The announcement came shortly after a report raised questions about CEO Sam Altman and the company’s approach to AI safety. In response, OpenAI is emphasizing compute access as a central part of the program, with fellows working directly alongside its internal safety teams.
The fellowship will run from September 2026 to February 2027, with total compensation exceeding $111,000. The company said it aims to attract external researchers and engineers to focus on key areas such as safety evaluation, robustness, ethics, privacy, and mitigating high-risk misuse.
The move comes as competition in AI safety research grows. Rival Anthropic already runs a similar fellowship with nearly identical benefits, including compute support and stipends.
The broader conversation around AI safety has intensified, especially following criticism highlighted in The New Yorker, which cited concerns from industry insiders about OpenAI’s past decisions, including the disbanding of its superalignment team.
As AI systems become more powerful, access to large-scale compute is increasingly seen as a key resource in both research and competition across the industry.