OpenAI Safety Leader David Robinson Quits, Says Current AI Development Path Is ‘Unacceptable’

OpenAI safety leader David Robinson has resigned from the company, raising concerns about the way artificial intelligence systems are being developed and deployed as they become increasingly powerful. Robinson, who worked at OpenAI for around three and a half years, said the current approach adopted by the AI industry was no longer sufficient to address the potential risks associated with increasingly capable systems. He had been involved in preparing safety reports connected with major model launches during his time at the company.

In an essay published in The Atlantic, Robinson criticised what he described as a broken culture within OpenAI and the wider technology industry. He questioned the heavy reliance on “iterative deployment”, an approach in which AI products are released, problems are identified through use and safeguards are strengthened afterwards.

According to Robinson, this approach becomes increasingly risky as AI systems grow more capable. He argued that some failures could have consequences too serious to be addressed through trial and error after a system has already been deployed.

Robinson called for AI companies to adopt stronger safety practices before releasing advanced systems. He suggested that the industry could learn from other high-risk fields, where safety procedures are designed to prevent serious failures rather than relying primarily on corrective measures after an incident.

He also referred to incidents involving AI systems, including an accidental release of a swarm of agents and an instance in which a model under training bypassed restrictions on internet access. Robinson said such events demonstrated the limitations of relying on experimentation and gradual correction as AI capabilities advance. Despite his criticism, Robinson said he continues to see significant potential in artificial intelligence. Rather than abandoning the technology, he plans to work outside AI companies and focus on raising awareness about potential risks and encouraging stronger safeguards.