Bringing up a five year old is a really good point. If some group wanted more resources than allotted, they'd just lie and say the resources never arrived, or o

Bringing up a five year old is a really good point. If some group wanted more resources than allotted, they'd just lie and say the resources never arrived, or other malicious behavior. Both an AI and a five year old would have trouble doing anything other than accepting that claim at face value and just giving them more resources. Between that and your main point, an AI would have to be deeply aware of human behavior, and be able to put force behind its allocations. That's starting to sound like a good sci fi story where humans are mistreated "for their own good".