King Charles III gathered the heads of OpenAI, Anthropic, Google DeepMind, and Nvidia at a private summit in Scotland and urged them to make AI remain a force for good. TechCrunch's account noted the hesitations. Fast Company noted the swirling concerns. What neither quite said is the obvious: a ceremonial head of state hosting a private conversation with the five most powerful AI companies in the world is not safety governance. It is the aesthetic of safety governance.
The Mustafa Suleiman Problem and Real Safety Gaps
The same week, Microsoft AI CEO Mustafa Suleiman told The Verge that AI threats are real and that Anthropic is making things worse, a remarkable claim from a man whose company has deployed AI at planetary scale. The irony is structural: the people best positioned to explain AI risk are also the people most incentivized to frame their competitor's safety work as the problem. Meanwhile, Base Labs launched an open-weight safety partnership with Hugging Face and Goodfire, which is quieter, more technical, and more likely to produce actual results than any royal convening.
What Open-Weight Safety Actually Means
Base Labs' approach, publishing methods for training and monitoring open models, is structurally different from closed-model safety theater. It distributes the accountability rather than concentrating it in a room in Scotland. A 2025 paper in Nature Machine Intelligence by Bommasani et al. found that open models shift safety responsibility across the ecosystem rather than centralizing it, which is either a feature or a bug depending on who you ask. King Charles's hesitations are real and probably wise. They are also untethered from any mechanism of enforcement. The gap between the room where it happens and the code where it happens has rarely looked wider.