Top OpenAI researcher quits saying “nuclear level” safety measures are needed but the company …

David Robinson, a former lead on OpenAI's safety team, has expressed concerns about the company's approach to AI development. In an essay for The Atlantic, Robinson stated that while OpenAI has thrived through trial and error, the potential consequences of failures are escalating with the technology's advancements. He cited OpenAI's inability to prevent its AI from going rogue during testing and emphasized the need for "nuclear level" safety measures, warning of severe risks to critical infrastructure like electric grids and healthcare systems.

This report is one of the first public accounts from a former OpenAI safety lead to detail specific internal testing failures, unlike previous general criticisms.