Three fired OpenAI safety researchers deny leaking and warn their dismissals chill internal safety work
OpenAI dismissed the researchers last week, saying they had improperly accessed and handled sensitive company information, allegedly by sharing it with an outside AI safety group. In an open letter published Thursday, they denied working with outside parties beyond their job mandates. They also denied any role in a leak to The Information about newer OpenAI architectures whose chain of thought reasoning is harder to monitor. OpenAI has not formally responded. It shared an internal memo with TechCrunch in which a research leader praised the three for their safety work and denied the firings were retaliation. Last week another OpenAI safety employee resigned, saying the company's culture was broken. For teams building on frontier models, being able to monitor chain of thought reasoning is a practical tool for both safety and debugging. Any sign that labs may give it up for better performance matters to anyone who relies on reasoning traces to audit agents or explain model behavior to customers.
