OpenAI’s Safety Fellowship Is Not a Generic Grant: It Trades Model Access for Structured, Governance-Focused Research
OpenAI’s Safety Fellowship, announced on April 6, 2026, is best understood as a structured safety research program rather than a broad grant round. It combines funding, mentorship, compute, and optional Berkeley workspace for independent researchers, but it withholds internal system access and sharply targets work that could inform AI governance, safety standards, and enterprise risk…