Announcing the OpenAI Safety Fellowship

Source: OpenAI Date: 2026-04-06 URL: https://openai.com/index/introducing-openai-safety-fellowship

Summary

OpenAI’s announcement of the OpenAI Safety Fellowship — a funded program for researchers to work on AI safety problems, similar in structure to the earlier Scholars program but specifically focused on safety research. The April 2026 launch follows the dissolution of the Superalignment team and represents OpenAI’s attempt to rebuild visible safety research infrastructure.

Implications

Safety/governance thread. The Safety Fellowship is a reboot of OpenAI’s external safety research engagement after the Superalignment team’s dissolution in 2024 created a credibility gap. The fellowship format (bring external researchers in, provide resources) is less committing than a dedicated internal team, but signals ongoing investment in safety as a research priority. The timing alongside GPT-5.5’s launch is notable — OpenAI is simultaneously pushing frontier capability and relaunching safety research infrastructure, positioning the two as compatible rather than in tension.