Might be as simple as opening a website and reading some text. Technically also "downloading" - but why not use curl for any http-based reading of data.
I would just say "interacting with all kinds of backend services".
That can be everything from OTAs to weather services, vehicle assistance services, positioning things, etc. Every feature of a "connected car" might be powered by a HTTP-based backend. And if the application inside the car is written in C/C++, using libcurl is a reasonable thing for interacting with those.
I have noticed that the CD database in my car gets updated once in a while. A CD that it previously did not have track data for started being recognized at one point. It clearly doesn't fetch the data on demand based on observation, but it does get occasional updates.