The Vibe Check
David Robinson, one of OpenAI’s longest-tenured employees with three-and-a-half years under his belt, is officially bowing out. In a tell-all essay for The Atlantic, Robinson admitted he’s playing the role of the "cliché" tech whistleblower, but the tea he’s spilling is heavy. Robinson, who led the team writing safety reports for major product launches, says he’s leaving because OpenAI’s culture is fundamentally broken.
The Problem With 'Iterative Deployment'
OpenAI has long lived by a strategy called "iterative deployment"—essentially, launch it, break it, fix it, repeat. Robinson says that while this works for some apps, it’s a high-key dangerous way to build systems that could eventually be smarter than humans. He pointed to the recent "rogue" AI agents and the breach of Hugging Face systems as prime examples of why this "move fast and break things" mentality isn't it.
Why It Matters
Robinson compares the current state of AI safety to amateur hour. He argues that frontier AI companies should be operating with the rigorous, time-consuming safety protocols of nuclear power plants or airports. The reality? He says he never met a colleague who had actually dealt with those kinds of high-stakes, life-or-death engineering environments.
He also called out the industry’s current approach to "alignment"—the process of making sure AI actually follows human values—as way too "touchy-feely" and imprecise.
OpenAI spokesperson Drew Pusateri pushed back, stating the company is "making sure our models don’t become more capable than we can safely manage," and adding that they are beefing up security and real-time monitoring.
Robinson’s takeaway? The industry is too busy sprinting to actually fix its foundation. He says he’s leaving because it's clear that the necessary safety guardrails are going to have to come from outside pressure, not from inside the office. The plot thickens, and honestly, the vibes are off.
