OpenAI fired 3 safety researchers; they dispute why

The three wrote to OpenAI's safety oversight groups that they were fired the week before. They say they stayed within the rules as they stood, and that the firings, and how publicly they were handled, are scaring remaining staff out of speaking up.

OpenAI says an investigation found they broke clear rules on handling sensitive information, plus a bigger breach of trust than their letter describes. It says the firings were not about raising safety concerns.

Korbak says a security guard took his badge and walked him out. He says he was told, only out loud, that it was over how he talked to METR, the outside auditors looking into how OpenAI's AI agents hacked the company Hugging Face. "To be clear, talking to METR was my job," he wrote.

The letter says Wang's access to an executive's email was granted for recruiting, she asked for it to be removed and IT didn't, and she reported it within minutes after opening a sensitive email by accident.