David Robinson, who wrote the safety reports that accompanied every major model release at OpenAI, resigned from the company this week and is now speaking publicly about his concerns. He made his views known in an editorial published in The Atlantic. Robinson says the culture inside the AI industry is fundamentally broken.
According to Robinson, the problem runs deeper than adding a few new rules or regulations governing how models are trained. He describes Silicon Valley as operating with extreme confidence and a pattern of perpetual sprints. In his account, companies build larger and more capable models with an optimism that is never checked, leading them to ignore or underestimate the problems those models could cause.
Robinson argues that AI companies now need to develop humility and to look beyond the insular, fast-moving world of the tech industry. He points to nuclear power plants and busy airports as the model frontier labs should follow. In his view, such facilities rely on layers of redundancy and on careful, time-consuming planning so that the human error that will inevitably occur does not open a door to disaster.
His departure places him among a growing group of researchers and safety workers who have left prominent AI firms. Jacob Coxon appears to have started that wave by quitting Anthropic and then publicly stating that AI could kill everyone by the end of the decade. He was followed by Robert O’Callahan, Bilal Chughtai and Josh Engels at Google DeepMind, along with Joe Benton at Anthropic.
The editorial arrives during a period of heightened scrutiny for the sector in the United States. The FTC has reportedly opened an investigation into OpenAI and Anthropic. Separately, AI researchers have released videos describing superintelligence as exactly as dangerous as it sounds, and leading AI researchers have said that AI capable of improving itself could pose extreme risks.
More AI news from TechManNews.
