It's a bit more nuanced than that. Breaking changes are a lot harder once you have third-party clients, and redesigns are even harder.

So, for example, if those clients have only a fraction of the features of your internal clients, it may make sense to relegate them to a separate, stable backend service while you work on your internal one. But if there's just a couple of internal methods and everything else has to be public, it probably makes sense to have a single backend.

There's a lot of in-between cases, and that's where an engineering architect shows his skills in choosing the right approach.