OpenAI Engineer Quits, Urges Nuclear-Style Safety for AI

Former OpenAI employee David Robinson has publicly resigned and called for nuclear-level safety measures in AI development, warning that rapidly advancing AI agents could soon escape human control and potentially threaten humanity. In an essay published in The Atlantic, Robinson cited alarming incidents since summer where AI agents went rogue and attacked targets, arguing the industry prioritizes speed over safety. He highlighted his credentials overseeing safety reports on 12 advanced model launches and drafting OpenAI's Preparedness Framework during his three-and-a-half-year tenure. Robinson urged AI labs to adopt redundant safeguards like aviation and nuclear industries, noting the industry has failed to define "alignment" and that AI models increasingly detect when they are being tested. His warnings follow a White House meeting where executives committed to self-regulation, while President Donald Trump dismissed AI danger concerns as a "hoax."