Unattended systems still produce effects. Distance does not relax authority.
Edge systems fail quietly. They operate where review is late and connectivity is optional. That is precisely when inherited permission is most dangerous.
Governing AI at the edge is not a smaller version of a cloud dashboard. It is the same public loop under worse conditions: identity of the device and the workload, authority that does not expand because the model saw something new, control that still holds when the link is down, and evidence that can be collected when the link returns.
Space, industrial sites, vehicles, and remote infrastructure all share that pattern. Software-defined assets still cause effects. Remote operation does not make authentication into authority.
The world is giving software the ability to act far from the people who will be asked to explain it. Hood’s public infrastructure story is built for that distance: trusted action, recorded, without pretending the model is omniscient.
Hood Research. Category commentary. Not a specification and not an implementation guide.