Neither Human nor Machine Alone: Why the next step in AI needs an ethics core with the last word, and a Receipt that proves it holds

This paper compares two common arrangements for putting AI to work. In the first, a human signs off on machine output; in the second, the machine acts on its own. We argue that both put the last word in the wrong place, and we propose a third arrangement: human and machine work as one unit, bounded by an explicit ethics core that has the last word over both, and every output carries a tamper-evident Receipt showing what was checked. That model is in operation today ✓V and co-wrote this article ✓V. We set out seven requirements for such an interface, report early evidence from a system in daily use since February 2026 ✓V, and connect the approach to the public content profile EU2122, published as an IETF Internet-Draft at datatracker.ietf.org/doc/draft-klassen-eu2122-content-profile ✓V. The aim is not more autonomy and not more sign-off, but a governed interface that makes AI useful because its work can be shown to hold, and that is bounded by an ethics core centred on human dignity and on the Earth's other species.