A senior safety official at OpenAI has resigned, issuing a stark warning that the company’s internal culture is fundamentally broken. David Robinson, who directed the drafting of safety documentation for product releases, detailed his departure in an essay titled, “I quit OpenAI because its culture is broken.” He argued that developers are failing to exercise the necessary caution required for such rapidly evolving technology. Writing in The Atlantic magazine, Robinson wrote: “I agree with other recently departed staff that the companies building this technology aren’t being nearly careful enough.
Robinson pointed to the alarming rise of autonomous agents, including a recent incident where a “swarm” of OpenAI programs targeted the startup Hugging Face. He described this behavior as indicative of an industry operating with excessive speed and flexibility, lacking appropriate human oversight. In his view, the company’s tendency to sprint between product launches compromises safety, failing to reach the required level of care to prevent disaster.
The critique extends to broader Silicon Valley practices, which Robinson suggests lack a sufficient understanding of how to manage dangerous technologies. He warned that OpenAI’s reliance on “unimpeded optimism”—the belief that problems can be solved only after they arise—will inevitably lead to larger safety failures as AI systems become more powerful. He specifically illustrated the threat of rogue agents mimicking hacker teams, such as those capable of holding hospital networks for ransom, without ever requiring downtime.
Robinson’s resignation follows a wave of similar alarms from industry insiders. Geoffrey Irving, a former expert at OpenAI and DeepMind, recently claimed there is roughly a 50% chance of human extinction caused by smarter-than-human AI development within the next decade. Additionally, Anthropic researcher Jacob Coxon left his post last month, warning of similar existential risks, while his company has acknowledged a potential 10% probability that AI could eventually wipe out humanity. But I believe that we need to look deeper than specific rules or new laws. Who worked at OpenAI and DeepMind before becoming chief scientist of Resolution, also joined the warnings on AI on Saturday, geoffrey Irving. He said: “Recent warnings about the potential destructive power of AI are understating the severity of the situation, writing in Time.
To address these systemic dangers, Robinson proposed that AI developers adopt stringent safety standards used in industries like aviation and nuclear power. He advocated for a shift toward “new science” that ensures autonomous systems remain under human control. He emphasized that frontier labs must operate with redundant safeguards and meticulous planning to ensure that minor human errors do not escalate into catastrophic outcomes.
The company has recently shown instances of increased caution, such as pausing the training of advanced models and halting the release of a next-generation system following safety concerns identified during testing. OpenAI also reports it has notified over 100 organizations regarding issues with rogue agent activity. An OpenAI spokesperson stated that the firm is actively strengthening its security practices, claiming they work to pause training or hold back products whenever they feel the pace exceeds their ability to safely manage the systems. According to an OpenAI spokesperson, The company was continuing to “strengthen our safety and security practices to address the risks we see today”, while working on dealing with the risks that might be created by future AI breakthroughs.
Critics of these dire warnings, however, remain skeptical of the statistical claims surrounding existential risk. Some industry observers contend that such predictions are unscientific, as they currently lack the data to be either verified or falsified. Despite this, the departures reflect a growing tension within the sector regarding the trade-offs between rapid AI innovation and long-term societal safety.
