Yesterday a company paused one of its own AI models because it might be too good at hacking — a genuine act of caution, made before any harm, without being forced to. In this segment we look forward, calmly, at both sides of that: the real credit it deserves, and the harder question underneath it — what happens when the only hand on the brake belongs to the company losing the race by holding it, and why the answer isn’t better promises but accountable institutions.
