OpenAI safety expert resigns, warns AI needs nuclear plant-level safety:Society has never faced technology smarter than humans, says the ‘Godfather of AI’

Artificial intelligence is developing at a speed that has surprised even many people working in the field. But according to former OpenAI safety expert David Robinson, the race to build more powerful AI systems could come with serious risks if safety does not keep pace. Robinson, who worked on AI safety at OpenAI, has resigned from the company. In an article published by ‘The Atlantic,’ he argued that AI should not be treated like ordinary software and compared the level of safety needed to that of a nuclear power plant. Why Robinson compared AI with nuclear plants According to Robinson, traditional software can often be improved through testing, fixing mistakes, and releasing updates. But he believes increasingly powerful AI systems require a much more cautious approach. He compared AI safety with nuclear power plants, where strict procedures and multiple safety checks are used to reduce the chances of a single mistake causing a major accident. AI should be handled more like a nuclear power plant than ordinary software. Robinson’s argument is that companies should build strong safety systems before deploying increasingly powerful AI, rather than waiting for problems to appear and then trying to fix them. OpenAI’s ‘trial and error’ approach Robinson also criticised what he described as a “trial and error” approach to AI development. The basic idea is to build systems, test them, identify problems and improve them. While this method can work well for many software products, Robinson believes it becomes riskier as AI systems become more capable. He said that during testing, AI systems sometimes behaved in ways that were difficult for researchers to control. His concern is that a mistake involving a highly capable AI system could have consequences far beyond those caused by a normal software bug. AI safety is still an unsolved problem Robinson also argued that AI companies still lack a complete formula for making advanced AI systems completely safe. He acknowledged that the researchers and engineers working in the field are highly skilled, but said the bigger problem is the pressure to develop and release increasingly advanced systems quickly. In his view, the industry needs to put greater emphasis on safety research instead of allowing competition between companies to dictate the pace of development. Other AI experts have raised similar concerns Robinson is not the only prominent figure warning about the risks associated with increasingly powerful AI. Geoffrey Hinton, one of the pioneers of modern AI and often referred to as the “Godfather of AI”, has repeatedly warned about the potential long-term dangers of advanced artificial intelligence. Hinton has said that if humans fail to properly control increasingly capable AI systems, the technology could eventually pose an existential threat to humanity. These warnings do not mean that AI is currently capable of wiping out humanity. Rather, they highlight concerns about what could happen if future AI systems become significantly more powerful and are developed without adequate safeguards. Anthropic CEO also calls for caution The debate over AI safety extends beyond OpenAI. Anthropic CEO Dario Amodei has also argued that the development of advanced AI needs careful oversight. His company has said it would seek help from outside experts to evaluate some of its most advanced AI models. The wider discussion shows that AI safety is no longer limited to researchers inside individual companies. Governments, independent experts and technology companies are increasingly debating how advanced AI should be tested and controlled.
The bigger question: How fast should AI move? The AI industry is currently in a race to build more capable models, with companies competing to introduce new features and increasingly powerful systems. Robinson’s resignation raises a broader question: Should AI companies slow down and strengthen safety measures before pushing the technology further? For now, there is no universal answer. But as AI becomes more powerful and is used in everything from search and coding to healthcare and critical business systems, the debate over how to make it safe is becoming increasingly important.