Ask HN: How do you manage prod data in dev for machine learning modelling?

We are on AWS and Machine lerning team wants to have production data lying in AWS production environment into our development account for their model training. Compliance is not allowing team do not allow cross access and we are left with the option of copying data from prod to a staging environment first and then to dev enviroment. How are you dealing with such situations. What are some best practices to adopt ?