They won't see the data of people they don't follow, and if they pop up and start following everyone they're going to get defederated.
Of course, it's relatively trivial to scrape the information without setting up a fediverse instance, so I don't think mastodon.prism.gov is something we'll need to worry about any time soon.
What sort of data are you talking about? I see posts and comments from people I don't follow in Local and All communities. Actually how do you follow an account in Lemmy?
So, on lemmy, you get content federated to you from remote instances by subscribing to communities on those remote instances. If a community exists on a remote instance, but no one from your instance subscribes to it, then no one on your instance will see any of the content posted to that community.
It works similar for mastodon and the regular fediverse, but in those cases, you follow people instead of communities.
Either way though, if no one from your instance subscribes to a particular remote account, be it a person or a community, then you don't get the content from that account on your instance.
Fun fact, this is because lemmy communities are the same type of entity as mastodon accounts. They're basically a "person" to the rest of the Fediverse.