OpenAI Safety Expert Resigns and Criticizes the Company

    David Robinson, who prepared the safety reports accompanying all major model releases at OpenAI, resigned this week. After leaving the company, he openly criticized the work culture that has emerged both at OpenAI and across the AI sector. According to Robinson, this approach prevents risks from being properly assessed.    Robinson believes there are serious problems with the work culture in the AI sector. According to him, companies are overly confident and are constantly trying to develop more powerful models. Such optimism causes potential dangers to be overlooked or made to seem less serious than they are.    According to the former OpenAI employee, merely adopting new rules is not enough. He believes leading AI laboratories should operate under strict safety rules like nuclear power plants and busy airports. A multilayered system of safeguards and careful planning should prevent human error from turning into a catastrophe.    Robinson is one of several experts who have recently left prominent AI companies and spoken about the dangers. After resigning from Anthropic, Jacob Coxon said that AI could kill all humans by the end of the decade. Robert O’Callahan, Bilal Chughtai and Josh Engels are other safety experts who left Google DeepMind, while Joe Benton left Anthropic.