Another OpenAI employee departs with a warning: 'parade of former colleagues' grows
David Robinson, who worked on safety issues at OpenAI, has left the company and warns in The Atlantic that AI companies' 'trial and error' approach must stop. He joins a growing group of (former) AI company employees publicly criticizing their former employers.
Another departure at OpenAI over safety concerns
David Robinson, who held a role in safety at OpenAI, announced his resignation last week. In doing so, he joins what he himself describes as a "parade of former colleagues": a series of employees at AI companies who have given up their positions due to concerns about the technology or the way it is being deployed. This is reported by Fast Company in an overview article about AI insiders speaking out publicly.
Exact details about Robinson's job title and the precise date of his departure have not been fully confirmed. What is clear, however, is that his departure fits a pattern that has been observed more frequently within the industry in recent years.
Criticism of the 'trial and error' approach
After leaving, Robinson wrote an essay for The Atlantic arguing that the 'trial and error' approach AI companies use in developing and rolling out their technology must stop. He is referring to the practice of launching AI systems and adjusting them based on problems that only come to light after release, rather than thoroughly understanding risks beforehand. The title and publication date of the essay are not further specified.
A recurring pattern in the industry
Robinson is not the first to leave a position at a major AI company with a public warning. Fast Company also mentions Ed Newton-Rex, who gave up his position in November 2023 over concerns about copyright and the use of training data by AI models. According to publicly available information, Newton-Rex was affiliated with Stability AI at the time, although this is not explicitly repeated in the material provided.
The full list of insiders mentioned by Fast Company in the article is not completely available. It presumably includes several names of (former) employees of major AI companies such as OpenAI, Google DeepMind, and Anthropic. In 2024, for example, members of OpenAI's so-called Superalignment team departed with similar criticism of the balance between rapid product development and safety safeguards, although it is not confirmed whether they are mentioned in this specific article.
Questions about governance remain
Robinson's departure feeds into the broader discussion about internal governance at AI companies: how much room is there for careful risk assessment when commercial pressure and competition are high? No response from OpenAI regarding Robinson's departure is known.
Looking ahead
The growing series of departing employees publicly voicing criticism points to ongoing tensions within the AI industry between speed of development and diligence. A complete picture of all the insiders mentioned and their specific criticisms requires further reporting; Fast Company and The Atlantic remain the central sources for further developments on this topic.