AI safety researcher says OpenAI ties compromised his integrity

An AI safety researcher says he joined OpenAI partly to help shape AI's future, but that fear of getting on executives' wrong side often governed his actions there. He says he made excuses for behavior that contradicted the company's stated goal of making artificial general intelligence go well for humanity. His criticism extends beyond his own conduct: he argues that researchers' reluctance to offend OpenAI executives helped AI companies capture the safety field, and alleges that fear of non-disparagement agreements kept people silent about company misconduct.