According to CoinMeta, David Robinson, the security chief of OpenAI, resigned due to a "breakdown" in the company culture. Over the past three and a half years, he was responsible for preparing security reports for 12 major model releases and published articles in The Atlantic criticizing the company's culture of trial and error. Robinson's resignation came shortly after OpenAI fired three employees related to security, who were dismissed for mishandling sensitive information. Robinson pointed out that OpenAI's "iterative deployment" model became more dangerous as the AI system gained more autonomy, as failures could become more severe before the company had time to correct them. He suggested that cutting-edge AI laboratories should be designed with multiple layers of redundancy, similar to nuclear power plants or busy airports, to prevent individual errors from escalating into major failures.