OpenAI Safety Exodus

THE PEOPLE PAID TO WORRY ARE WALKING OUT.

OpenAI fired three safety researchers for allegedly sharing confidential information with an outside AI safety group, then days later its safety lead David Robinson resigned after 3.5 years, calling the company’s culture “broken” in an Atlantic essay declaring “the time for trial and error is over.” Meanwhile Sam Altman told Reuters the world should accept “some bad things happening” in exchange for AI’s benefits. This is the tracker for OpenAI’s internal safety crisis: who left, what they said, and who got fired.

1. Oct 1, 2026

   OPENAI FIRES THREE SAFETY RESEARCHERS FOR ALLEGEDLY LEAKING TO AN OUTSIDE SAFETY GROUP

   OpenAI parted ways with three safety researchers it accused of sharing confidential information with an outside AI safety group, the Wall Street Journal reported on October 1, 2026. The stated cause was the researchers’ contact with the safety group — not their research — and the firings landed days before the company’s safety lead resigned. The episode reframed OpenAI’s safety work as an information-security fight.

   SOURCES: The Wall Street Journalreported

   

   THE SAFETY EXIT

2. Oct 3, 2026

   SAFETY LEAD DAVID ROBINSON QUITS: THE CULTURE IS “BROKEN”

   David Robinson, OpenAI’s safety lead, resigned on October 3, 2026 after 3.5 years and 12 published safety reports, and published an Atlantic essay calling the company’s culture “broken.” His central line — “the time for trial and error is over” — became the slogan of the exodus. His exit came days after the company fired three safety researchers for allegedly leaking to an outside safety group.

   SOURCES: Reutersreported · TechCrunchreported · Startup Fortunereported

   

   THE LAST CONE

   

   THE LAST SAFETY CONE

3. Oct 4, 2026

   ALTMAN: THE WORLD SHOULD ACCEPT “SOME BAD THINGS HAPPENING”

   On October 4, 2026, the same weekend his safety lead quit, Sam Altman told Reuters that AI’s benefits warrant accepting “some bad things happening” — the company’s risks framed as the price of progress. The line landed while the safety exodus was still front-page news, and became Exhibit A for the departing researchers’ argument that caution was losing the internal fight.

   SOURCES: Reutersreported

   

   THE APOLOGY TOUR

   

   THE WHISTLEBLOWER UNION

MORE FILES

• OpenAIPALACE INTRIGUE. FUTURE BEFORE LUNCH.
• Sam AltmanFIRED. BACK IN FIVE DAYS.
• LASSTTAKING THE MACHINES TO COURT.
• FTCSIGNED THE SELF-POLICING ACCORD. GETTING SUBPOENAED ANYWAY.
• Super IntelligenceWASHINGTON RENAMED AI. THE ROBOTS DIDN'T NOTICE.
• Frontier Lab AlarmsTHE LABS ARE REHEARSING THE DAY AFTER.