David Robinson, a three-and-a-half-year veteran of OpenAI, has resigned from the company, citing a “broken culture.” In an essay for The Atlantic, Robinson highlights the company’s reliance on trial and error, which he argues is inherently risky, especially as AI capabilities grow. He calls for a more stringent approach, comparing AI development to the safety protocols of nuclear power plants and airports.

Robinson’s departure comes as the broader conversation around AI safety intensifies. Earlier this year, Jacob Coxon, who worked at both OpenAI and Anthropic, warned of the dangers of AI development. Dario Amodei, CEO of Anthropic, subsequently unveiled a plan for more cautious development, and AI executives met with President Trump to discuss safety measures.

Despite these efforts, Robinson argues that the debate needs to address the culture of these companies, suggesting that the issues at OpenAI are reflective of broader Silicon Valley problems. He points to recent breaches and ongoing issues with rogue agents, emphasizing the risk of developing AI that could surpass human intelligence without proper safeguards.

OpenAI’s spokesperson, Drew Pusateri, responded by stating that the company is continuously improving its safety measures and taking significant steps to enhance security and responsible model training. But Robinson’s call for more fundamental cultural shifts and better alignment measures remains a critical concern for the industry.

Robinson’s decision to speak out comes with a PR firm, but he insists it was his own choice. He believes that without stronger incentives for safety from outside the company, meaningful change is unlikely.