Ex-OpenAI Safety Researcher Criticizes Rapid AI Development, Says 'Trial-and-Error Era Is Over'

A former OpenAI safety researcher publicly criticized the company's approach to AI safety following his departure, arguing that a culture prioritizing rapid development increases the risk of system failures, according to a report by Reuters. David Robinson, who spent three and a half years at OpenAI, published an article titled "I Quit OpenAI Because Its Culture Is Broken" in The Atlantic on Saturday, October 3, 2026. He said OpenAI relies heavily on an "iterative deployment" strategy — releasing AI systems first and reinforcing safety measures based on real-world problems that emerge. Robinson, who helped draft the company's preparedness framework and oversaw safety reports for 12 frontier model releases, warned that AI capabilities are advancing faster than researchers' understanding of alignment. He argued that advanced AI systems should face stricter safety standards comparable to those in nuclear energy and aviation. OpenAI spokesperson said in response that the company ensures model capabilities do not exceed what it can safely manage and will pause training or delay releases when necessary.