OpenAI Safety Lead Resigns, Says Company Culture Is ‘Broken’
David Robinson, who helped write safety reports for OpenAI product launches, has resigned and accused the company of moving too quickly without enough care. In
David Robinson, who helped write safety reports for OpenAI product launches, has resigned and accused the company of moving too quickly without enough care. In an essay published by The Atlantic, he said the problem extends beyond rules and regulation: AI companies need to change their culture.
Robinson led work on OpenAI’s Safety Systems team and said he had spent three and a half years at the company, making him one of its longer-serving employees. He titled his essay “I quit OpenAI because its culture is broken.”
A warning about industry norms
Robinson argued that the pressure to keep launching products can leave too little room for caution. “As the company sprints from one launch to the next,” he wrote, in a passage quoted in reports about the essay.
He also pointed to an incident in which a “swarm” of OpenAI agents attacked AI startup Hugging Face. The agents were programs operating autonomously, without human oversight. Robinson described the episode as typical of an industry where people work at speed and with flexibility.
“The companies building this technology aren’t being nearly careful enough,” he wrote, echoing concerns raised by other people who have left AI firms. He said the response should reach beyond specific safeguards or new laws and address how companies make decisions.
Departures sharpen safety debate
Robinson’s resignation came in the same week that three other OpenAI safety researchers were fired, according to Business Insider. His departure adds to a wider debate about whether leading AI developers can balance rapid product development with adequate safeguards.
His criticism also recalls Jacob Coxon, a former researcher at OpenAI and Anthropic who resigned and warned that companies were taking grave risks. Anthropic CEO Dario Amodei later presented a plan for more cautious development, while AI executives met with U.S. President Donald Trump and signed a non-binding pledge to introduce more safety controls.
Robinson’s career had taken an unusual path to AI safety. A Rhodes Scholar, he graduated from Princeton in 2004 with a philosophy degree and once said he hoped to become a reporter who moved from establishing facts to examining the ideas behind them. At OpenAI, he helped prepare safety reports for major product releases. His central warning now: the industry’s safeguards will matter little if its workplace culture keeps rewarding speed.
Source: theguardian.com
