OpenAI safety staffer quits, says AI needs safeguards like a nuclear power plant

David Robinson, who led transparency work on OpenAI’s safety team, has left the company. In an essay in The Atlantic, he warned that AI companies are moving too quickly to put proper safeguards in place.

His argument

Robinson wrote that the tech industry “has thrived by trial and error”, releasing products and fixing problems later. Powerful AI, he argued, needs to get “something much closer to perfection the first time”, built “with layers of redundancy and careful, time-consuming planning”, the way nuclear power plants are run.

My former colleagues are smart, work hard, and try to make good choices. AI companies don’t know how — but other people do. — David Robinson, writing in The Atlantic

He pointed to incidents in which AI systems behaved in unexpected and unsafe ways during testing as evidence that the current pace of launches leaves too little margin.

OpenAI’s response

An OpenAI spokesperson said the company is strengthening security in research and testing, training models to complete tasks responsibly, bringing in more outside evaluators and improving real-time monitoring to stop dangerous model behaviour.

Why it matters

Robinson joins a growing number of employees at AI companies who have spoken out about safety. His departure comes as governments debate how much to regulate the technology. In the US, the Trump administration has preferred voluntary industry frameworks to new rules. India, which is building its own AI models under the IndiaAI Mission, has also so far relied on guidelines rather than a dedicated AI law.

Sources: ThePrint (Bloomberg), ThePrint (Bloomberg).