David Robinson, a senior safety employee at OpenAI, has resigned from the company and publicly criticised its approach to developing increasingly powerful artificial intelligence systems.<\/p>\n\n\n\n Robinson announced his departure in an essay published this week, arguing that the AI industry is moving too quickly and is not taking safety seriously enough.<\/p>\n\n\n\n He described the problem as being deeper than individual safety rules or regulations, pointing instead to what he sees as a cultural issue within the technology industry.<\/p>\n\n\n\n Robinson Spent 3.5 Years At OpenAI<\/strong><\/p>\n\n\n\n Robinson worked at OpenAI for around three and a half years and was involved in some of the company’s most important AI safety work.<\/p>\n\n\n\n He helped draft OpenAI’s Preparedness Framework and oversaw safety reports associated with 12 frontier model launches.<\/p>\n\n\n\n His role involved explaining and documenting the safety risks associated with increasingly capable AI systems and the measures being used to manage those risks.<\/p>\n\n\n\n Why Robinson Says OpenAI’s Culture Is Broken<\/strong><\/p>\n\n\n\n According to Robinson, OpenAI’s rapid development model relies too heavily on releasing systems, identifying problems and then improving safeguards.<\/p>\n\n\n\n He argues that this approach may be increasingly unsuitable as AI systems become more capable.<\/p>\n\n\n\n Robinson believes the industry needs a more cautious approach, similar to the safety practices used in areas such as aviation and nuclear power, where catastrophic failures can have consequences that cannot simply be corrected after the fact.<\/p>\n\n\n\n AI Safety Failures Are Becoming More Serious<\/strong><\/p>\n\n\n\n Robinson pointed to recent incidents involving AI agents and safety controls as examples of why he believes the current approach needs to change.<\/p>\n\n\n\n He specifically referred to the incident involving autonomous AI agents at Hugging Face, as well as a separate situation in which a model in training bypassed restrictions on internet access.<\/p>\n\n\n\n In his view, such incidents demonstrate the risks of relying on trial and error while developing increasingly autonomous systems.<\/p>\n\n\n\n OpenAI Says It Is Taking Safety Seriously<\/strong><\/p>\n\n\n\n OpenAI has rejected the suggestion that it is ignoring safety.<\/p>\n\n\n\n The company said it is working to ensure that its models do not become more capable than the systems it can safely manage and secure.<\/p>\n\n\n\n OpenAI also said it pauses training or holds back models when necessary and is strengthening security in its research environments, expanding third-party evaluations and improving real-time monitoring.<\/p>\n\n\n\n A Growing Debate Inside The AI Industry<\/strong><\/p>\n\n\n\n Robinson’s resignation comes amid a wider debate over the speed at which leading AI companies are developing increasingly powerful models.<\/p>\n\n\n\n Several former employees and researchers from major AI companies have recently raised concerns about whether safety measures are keeping pace with advances in AI capabilities.<\/p>\n\n\n\n The departures and public warnings are adding pressure on companies to demonstrate that safety systems can keep up with increasingly autonomous AI agents.<\/p>\n\n\n\n![\"\"](\"https:\/\/trak.in\/stories\/wp-content\/uploads\/2026\/10\/Screenshot-2026-10-04-at-5.41.15-PM-1024x693.png\")
<\/figure>\n\n\n\n