Realistically, I think you're more hitting up on how to convey, than what the actual is. Probably this should be more dynamic, where different audiences receive the same, but differently phrased message.
Here's an example. Discrimination. Here's a factual phrase: women are less intelligent than men. That's not discriminatory, what is discriminatory, is if I use a minor deviation in averages, to look at $random_woman, and presume she is less intelligent than $random_man.
For example, there are many women that are more intelligent than many men.
My point is, we've had to "dumb down" discrimination, by trying to claim, in many respects, there there is absolutely no difference between men and women. This has led to detrimental outcomes for women, a key example being that as most medical volunteers are men, many models of the brain are predicated upon the male, and therefore? As people say "why, there's no difference between the male and female brain!", it has actually harmed women's outcomes.
I am concerned about the parallel here. That the more we lump things into mega-categories, the more positive outcomes for nuanced groups is lost. And that means that the best solutions may be unavailable, for if they cannot be applied to "overall group", then they may be deemed useless.
I know you're not advocating any of this. And I know your goal is to just ensure people fully understand the risks.
But my point is, much as the fight against discrimination has been whittled down to "there is no difference!", because the stupids cannot understand nuance, you seem to be in a trap of having to whittle down to no-nuance, for the same reasons.
Which is sad.