Former OpenAI safety staffer David Robinson writes that the company's culture is broken, and says AI labs should be run like nuclear plants

On Saturday 3 October The Atlantic published an essay by David Robinson titled “I Quit OpenAI Because Its Culture Is Broken”. His departure ran here on 3 October; this is his own account of why. The Atlantic's page could not be opened from here, so the essay is reported through CTech, ThePrint and DW.

CTech reports that over three and a half years Robinson helped draft the company's preparedness framework and oversee safety reports for 12 frontier-model launches. ThePrint describes him as having previously led transparency work on OpenAI's safety team.

His target is the way OpenAI ships: release, see what goes wrong, then add safeguards. In ThePrint's account he writes that the company has thrived by trial and error, and argues that this is no longer good enough as models grow more capable. “The time for trial and error is over,” he wrote, according to CTech. “As the company sprints from one launch to the next, it is failing to achieve the level of care that I believe is needed,” he added. He wants AI companies run more like nuclear power plants, “with layers of redundancy and careful, time-consuming planning, so that the occasional and inevitable human error does not open a door to disaster.”

An OpenAI spokesperson said in a statement quoted by CTech: “We're making sure our models don't become more capable than we can safely manage and secure, and we pause training or hold back models when we need to slow down”. ThePrint adds that OpenAI says it is strengthening security in its research and testing environments, expanding work with third-party evaluators and improving real-time monitoring.

A man who helped write OpenAI's safety reports now says the process behind them is not careful enough.

• Reported Robinson's Atlantic essay “I Quit OpenAI Because Its Culture Is Broken” was published on Saturday 3 October; he helped oversee safety reports for 12 frontier-model launches. CTech (Calcalist)
• Claimed Robinson says AI companies should be run more like nuclear power plants, with layers of redundancy and careful planning. ThePrint
• Claimed OpenAI says it pauses training or holds back models when it needs to slow down. OpenAI spokesperson, via CTech