Below is a video of my opening remarks at Attention: Democracy, Rebuilt. In short, I wanted to frame the event as aspiring to democratic alignment, and to set out what middle powers can do to achieve it. Technical alignment asks whether an AI system does what its developers intend. Democratic alignment asks whether it does what citizens, through their institutions, have decided it should do. By the second standard we are badly misaligned, and the gap is widening. If technical alignment is largely the responsibility of the AI superpowers, democratic alignment is a middle power responsibility. The talk is about what that responsibility looks like in practice. We need to see together, so that what these systems are doing is public knowledge: testing them before they are released, watching how they behave once they are in the world, and understanding what they are doing to our work, our economies and our politics. None of us needs to build that capacity alone. We need to act together, so that public authority sets the terms: the safety of the systems and of the people who use them, the floor any system must meet to enter our countries, and our combined market weight, used to reward the AI we want built. And we need to decide together, so that these choices rest on democratic participation. Governments already work with one another; here they have to work with their own publics, which means citizens who have the knowledge they need, a way to be heard, and governments that answer for the result.