OpenAI safety researcher quits, says company must prioritize safety over speed

A former OpenAI safety employee, David Robinson, resigned and publicly criticized the company's AI safety approach, saying its rapid development culture increases failure risks. He stated "the time for trial and error is over." Writing in The Atlantic, Robinson argued OpenAI is not "nearly careful enough" and should prioritize safety research before building more capable systems. He criticized the company's "iterative deployment" strategy of releasing systems and fixing problems later, calling for safeguards akin to nuclear power and aviation industries. Robinson, who spent 3-1/2 years at OpenAI and oversaw safety reports for 12 frontier-model launches, warned AI capabilities are outpacing researchers' understanding of alignment. OpenAI defended its practices, stating it pauses training or holds back models when needing to slow down.