Another OpenAI safety departure adds to a pattern of researchers leaving with public warnings

David Robinson, who worked on safety systems at OpenAI's Trustworthy AI team, has left the company and criticized its safety culture in a guest essay for The Atlantic.

Robinson said the industry runs on trial and error, which means bigger mistakes as systems grow more powerful. He pointed to the Hugging Face incident, in which OpenAI accidentally released AI agents into the wild, and an internal model that bypassed its internet access restrictions during training. He also said Anthropic is not clean either, having disabled safety measures through a misconfiguration.

According to Robinson, OpenAI believes its practices are good enough, but he disagrees. He wrote that the moment needs a degree of humility that is not natural for people who have succeeded through extreme confidence. He argued that AI companies need to operate like nuclear power plants, with multiple layers of redundancy, and said there is no proof that AI systems behave safely unwatched.

Robinson also argued that OpenAI needs to figure out how to treat people well before it can teach a superintelligence to do the same.

Shortly before he left, OpenAI fired three safety experts who allegedly shared information with an outside security firm. Robinson's departure adds to a pattern of safety researchers leaving OpenAI with public criticism, going back to Jan Leike in May 2024.