October 4, 2026, (Inside AI) — A former OpenAI safety employee who resigned recently has publicly criticized the company's safety culture, warning that its rapid development pace heightens the risk of catastrophic failures. David Robinson, who spent three and a half years at OpenAI and helped draft its preparedness framework, made the claims in an essay titled I Quit OpenAI Because Its Culture Is Broken, published by The Atlantic on Saturday.
Robinson's departure and his decision to speak out add a prominent voice to a growing industry debate over whether leading AI labs are prioritizing speed over safety. His critique centers on OpenAI's reliance on "iterative deployment," a strategy of releasing systems and then strengthening safeguards as problems emerge. Robinson argues that this approach is no longer adequate for advanced AI, which he says demands protections similar to those in nuclear power and aviation.
"The time for trial and error is over," Robinson wrote, contending that companies like OpenAI are not being "nearly careful enough" and should invest more in safety expertise and research before building more capable systems.
During his tenure, Robinson oversaw safety reports for 12 frontier-model launches, giving him a rare vantage point on the company's internal processes. He wrote that "as the company sprints from one launch to the next, it is failing to achieve the level of care that I believe is needed."
OpenAI pushed back on the characterization. A spokesperson said in a statement: "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."
Robinson also warned that AI capabilities are advancing faster than researchers' understanding of alignment, the field focused on ensuring AI systems act in accordance with human goals and values. That gap, he suggests, makes the iterative approach increasingly risky.
The criticism lands amid heightened scrutiny of AI safety practices. OpenAI and rival Anthropic have both faced incidents where safety controls failed or experimental systems behaved unexpectedly. Such episodes have fueled calls for more rigorous oversight and slower deployment.
Robinson's essay echoes concerns raised by former employees and outside experts who argue that commercial pressures at AI labs can undermine safety commitments. His call for nuclear- and aviation-style safeguards implies a regulatory framework with strict testing, licensing, and redundancy requirements, far beyond current voluntary guidelines.
OpenAI has long maintained that iterative deployment allows it to learn from real-world use and improve safety over time. But critics say that approach externalizes risk onto users and society. The debate is likely to intensify as AI systems grow more powerful and as governments consider new regulations.
For now, Robinson's resignation and his blunt assessment serve as a reminder that internal dissent within AI companies is becoming more public. Whether it leads to meaningful change in OpenAI's culture or the broader industry remains an open question.
