OpenAI Safety Lead Resignation
OpenAI's safety lead walked out, and he said the quiet part out loud.
David Robinson spent about three and a half years there writing the safety reports for major model launches. He led that work for 12 of them. His reason for leaving is blunt: the culture is "broken."
He is not objecting to one bad release. His argument is about the playbook. Ship, observe, patch. He compared that approach to nuclear plants and airports, where the industry does not learn from failures in production.
I don't read this as a villain story. Many people inside AI labs share the speed worry. What stands out is that Robinson's job was to write down the risks, and he has concluded the process can't carry them anymore.
He is reportedly the fourth safety staffer to leave in about a week, after OpenAI parted ways with three researchers over how they handled sensitive information.
If your AI vendor's safety team quit tomorrow, would you hear about it?