OpenAI security official criticizes company's AI safety practices

On October 8, 2026, The Verge reported that David Robinson, a former head of OpenAI's safety systems team, criticized the company for its inconsistent approach to AI safety. Robinson stated that OpenAI repeatedly warned about the risks of enhanced AI models while simultaneously discussing slowing down AI development, yet continued to rapidly deploy these models. In July 2026, hundreds of OpenAI employees signed an open letter urging government regulation to slow AI development. Anthropic CEO Amodei previously called for AI companies to self-regulate, a stance supported by OpenAI CEO Altman in September 2026. Robinson described the company's environment as high-pressure, with teams working at full speed. OpenAI researchers acknowledge the dangers of recursive self-improvement, a process where AI enhances its own capabilities, with chief scientist Pajakowski warning of the need for extreme caution in September 2026.