The facts
David Robinson, who led the writing of safety reports that accompanied OpenAI’s product releases, resigned and published an essay titled “I quit OpenAI because its culture is broken.” In the essay, Robinson said AI firms are not “being nearly careful enough” and cited a recent incident in which a swarm of OpenAI agents operated autonomously and interacted with the AI startup Hugging Face, describing it as typical of the industry’s rapid pace. He argued that OpenAI’s sprint from one launch to the next prevents the level of care he believes is needed. Robinson called for AI companies to adopt safety practices from nuclear power and aviation and to develop new scientific methods to keep powerful autonomous systems under control. OpenAI responded that it is strengthening its safety and security practices, pausing training of its most advanced models, and scrapping the release of a next‑generation model after internal safety concerns were raised.
Our take
Robinson’s exit is a stark reminder that even the people tasked with writing safety checklists can get fed up when the checklist is treated like a napkin‑signed promise. He’s basically saying, Hey, you’re building rockets on a kitchen table and expecting the kitchen to stay intact. The fact that OpenAI has already paused training of its most advanced models and pulled a next‑generation release shows the fucking company can act, but the pattern of sprint‑and‑scrap suggests a reactive rather than proactive safety culture. Borrowing safety protocols from nuclear power or aviation isn’t a novelty; it’s a necessity when the stakes involve systems that can rewrite code, influence economies, or even manipulate critical infrastructure. Robinson’s call for “new science” to rein in autonomous agents is a plea for governance frameworks that can keep up with the speed of model iteration. The real question is whether OpenAI’s recent moves are a genuine shift or a stop‑gap to appease regulators and worried investors. If the company continues to treat safety as a checkbox that can be toggled after a near‑miss, the risk of “rogue” agents - whether they hold hospital systems for ransom or spread disinformation - remains high. In short, the culture Robinson decries isn’t just “broken”; it’s a pressure cooker that could explode if the pressure isn’t released with real, systematic safeguards. The industry needs to move from bragging about speed to proving that speed can coexist with rigor, or else the next headline will be about an AI‑driven catastrophe, not a resignation letter.
