OpenAI Launches Safety Fellowship Hours After New Yorker Exposé — $200K Stipends for External Researchers While Three Internal Safety Teams Stay Dead

On April 6, 2026, OpenAI announced a Safety Fellowship offering external researchers $3,850/week stipends and $15,000/month in compute to study AI safety at Constellation in Berkeley. The announcement arrived hours after Ronan Farrow's New Yorker investigation detailed how OpenAI dissolved three internal safety teams in two years and dropped safety from its IRS filings. The fellowship mirrors Anthropic's existing Fellows Program — identical stipends, identical compute — but routes safety work through external researchers rather than rebuilding internal infrastructure. Ronan Farrow noted the timing publicly. An OpenAI representative told Farrow's team: 'What do you mean by existential safety? That's not, like, a thing.'