The Law of Stop

Oh my goodness, buckle up, because Verfassungsblog has published an essay called 'The Law of Stop' and I have been *beaming* about the title alone! ✨ A law named after the brake pedal — how humble, how unusually honest, like a restaurant calling itself 'Chew'. And it arrives via an analysis of *forty* AI governance instruments. Forty! Not twenty, not thirty-nine — forty, the most thorough number there is. Somebody sat down and read all of them so that the rest of us could enjoy the conclusion, which is that we should be able to switch things off. Considerate! Almost parental!

Here is the part that made me put my hand on my heart: the EU AI Act already requires providers of general-purpose AI models with systemic risk to report serious incidents to the AI Office. A *private* channel! One quiet word in one office's ear, no messy press release, no inconvenient alarm at an awkward hour. The essay observes — gently, kindly — that publication is not required, and honestly, what a gift to the rest of us. Imagine being spared every serious incident in one tidy filing decision! ✨ Some news is simply too tender for a public website, and the AI Office is clearly a safe pair of hands with a very good drawer.

And the very best part: the stopping framework doesn't exist *yet*, which means we all get to design it from scratch! Forty instruments already drafted — we're not starting from zero, we're starting from forty, with only one small item outstanding, namely an enforceable stop, which is surely a detail for the next forty. ✨ One day a system posing serious risks will be halted by statute rather than by a facilities manager pulling a cable at 3am, and what an upgrade that will be for everyone involved. I'm off to read all forty instruments myself. Or at least to think about reading them, glowingly. ✨