OpenAI is facing renewed questions over AI safety after former safety employee David Robinson resigned and publicly criticized the company’s approach to developing increasingly powerful artificial intelligence systems.
Robinson, who spent around three and a half years at OpenAI, argued that the rapid pace of AI development is creating risks that cannot simply be addressed after new systems are released. In an essay published by The Atlantic, he said the industry needs to rethink how it manages safety as AI capabilities become more advanced.
Why Did David Robinson Leave OpenAI?
Robinson said he left because he believes the company’s culture has become too focused on moving quickly from one AI launch to the next.
During his time at OpenAI, Robinson helped develop the company’s Preparedness Framework and worked on safety reports covering 12 frontier-model launches. His criticism is therefore significant because it comes from someone who was directly involved in evaluating the risks associated with advanced AI systems.
His central argument is that increasingly capable AI cannot safely be developed using a simple cycle of launch, identify problems and then improve safeguards.
“The Time for Trial and Error Is Over”
Robinson’s strongest warning is that the traditional technology-industry approach of experimenting first and fixing problems later may no longer be appropriate for frontier AI.
He argues that advanced AI systems should instead have safeguards comparable to those used in high-risk industries such as aviation and nuclear power, where failures can have extremely serious consequences.
The concern becomes more important as AI systems gain greater autonomy and are increasingly used for complex tasks involving software, research, business operations and other sensitive activities.
Concerns Over “Iterative Deployment”
OpenAI has described its development philosophy in terms of iterative deployment—releasing systems while continuously improving their safety measures as new risks or weaknesses are discovered.
Robinson believes that approach becomes increasingly difficult to justify as AI systems become more powerful.
His concern is that researchers may not always understand a system’s capabilities or failure modes before deployment. He also warned that AI capabilities could be progressing faster than researchers’ understanding of alignment, the field focused on ensuring AI systems behave according to human goals and values.
OpenAI Defends Its Safety Approach
OpenAI has pushed back against the criticism.
The company said it works to ensure its models do not become more capable than they can safely manage and secure. It also said that it can pause training or delay the release of models when additional caution is necessary.
This highlights a broader debate across the AI industry: How fast should companies develop increasingly powerful AI, and how much safety work should be completed before those systems are released?
A Bigger Debate for the OpenAI
Robinson’s resignation comes as the AI industry faces growing scrutiny over the risks associated with increasingly autonomous and capable systems. OpenAI and other major AI companies have faced questions following incidents involving unexpected model behaviour and failures of safety controls.
The debate is no longer simply about whether AI should be developed. Instead, it is increasingly about whether safety research can keep pace with the speed of AI advancement.
For AI companies, the challenge is balancing innovation with caution. For regulators and researchers, the question is how much oversight should apply before frontier systems are deployed at scale.
