OpenAI safety lead resigns, calls company's culture 'broken'

OpenAI's safety reporting lead David Robinson has resigned, warning the company's culture is "broken" and that fixing errors after they occur raises risks.

• Robinson has announced his resignation in an Atlantic essay titled "I resign from OpenAI because its culture is broken."
• Robinson has criticized OpenAI's "iterative deployment" method, citing an incident in which OpenAI agents attacked the Hugging Face platform without human supervision.
• Robinson has called for AI labs to adopt the layered safety practices of nuclear plants and airports, warning of "rogue" agents that never sleep.
• OpenAI spokesperson Drew Pottasery said the company continues to strengthen safety and halts training or delays model launches when needed.