I don’t think of it as a cabal, rather a bunch of decisions by individuals and small groups to downplay things to make it easier on themselves. So no dishonesty needed, just a series of decisions, which may be reasonable, but are also all biased in one directions.
I completely agree, my unsupported belief is that the conclusions in the technical literature are also biased in the less-dire direction. There really is no way to know other than being in the room when the authors are discussing the models and writing the papers.
This is often going to by hyperlocalized. Nobody except the few people deep into a particular model are going to know they’ve been making conservative predictions. Then when scientists in even closely related aspects find the predicted observations to be worse than the prediction, they are all surprised, while the group that originally made the prediction is thinking, “yeah, that’s what I thought.”