OpenAI safety employee quits, says 'time for trial and error is over' - Reuters

A former member of OpenAI’s safety team has resigned and published an essay titled “I Quit OpenAI Because Its Culture Is Broken,” in which he argues that the company’s development approach is insufficiently cautious. The essay, released through The Atlantic on Saturday, contends that the period of trial and error that has characterized the firm’s rapid rollout of new models must end.

The author, David Robinson, spent three and a half years at OpenAI, during which he helped draft the organization’s preparedness framework and oversaw safety reporting for twelve frontier‑model launches. His responsibilities included leading transparency work on the safety team and contributing to the evaluation of advanced AI systems before they were released to the public.

Robinson criticises OpenAI’s reliance on “iterative deployment,” a process that involves releasing systems and strengthening safeguards only after problems emerge. He argues that as AI capabilities grow, the stakes of inevitable failures increase, and that safeguards should resemble those used in high‑risk industries such as nuclear power and aviation. He says the company’s “sprint” from one launch to the next prevents it from achieving the level of care he believes is required, calling for a standard of “much closer to perfection the first time.”

His comments come amid a broader debate within the artificial‑intelligence sector about the speed of development and the adequacy of safety controls. OpenAI and rival Anthropic have faced scrutiny after incidents in which safety mechanisms failed or experimental systems behaved unexpectedly. The concerns raised by Robinson add to a series of recent warnings from employees across leading AI firms about the potential for catastrophic outcomes.

In response to the growing unease, Anthropic’s chief executive, Dario Amodei, has called for a slowdown in the development of the most advanced models and announced plans to involve third‑party evaluators in the assessment process. Other industry voices have echoed the call for more rigorous oversight, noting that some employees have warned the technology could eventually pose existential threats.

Robinson’s resignation and public critique highlight an internal push for stronger safety practices as AI systems become more powerful, underscoring the tension between rapid innovation and the need for robust risk mitigation within the sector.