"It should be trivial for 99% of companies to anonymize and scramble user data so that employees do not have access to it. There's no reason for it."

It isn't: https://en.wikipedia.org/wiki/De-anonymization

I advocate for in practice behaving as if it is not possible, and simply saying that production data must be restricted. Exceptions can be made for certain types of aggregate data, but one of the lessons of a lot of people's work is that it's really, really easy to cross the line back to the data being identifiable again without you even knowing you did it.

Also, I don't mean that wikipedia link as a complete explanation per se, but the starting point for a journey, if you want to take it.