I think this situation might be similar to google/images rather than youtube as others have suggested. Google certainly indexes tons of images which infringe on someone's copyright. Furthermore you can view them directly in google/images.
When you bill your service as a "streaming music service" 99.5% of your users will use it to listen to Taylor Swift, and by billing it as such single apparent inference is that you intend for them to do that. The non-infringing uses are just a fig leaf--something you point to trying to cover up the obvious.
Google Image search is different. People don't primarily use GIS to find images that are hosted on the internet without the copyright owner's permission, nor does Google do anything to market to that use.
I agree with your general interpretation of the case here, but I wouldn't be surprised if the primary use case of GIS actually is copyright infringement: finding images to paste into your Web site, PowerPoint presentation, Word document, or what have you. Google is, interestingly enough, rather silent about what GIS is actually for.
Related is YouTube, which even post-lawsuit-and-Content-ID gets a huge amount of use for music piracy. But it's probably easier for them to argue post-Content-ID that they've done a lot to actively discourage that use.
Compare it to Tor. Lots of legal uses (at least legal in the US, maybe not in the country where one would need something like Tor for said uses). But the majority of its use is for a few illegal activities. Are we to judge intent based off of how the majority of uses use it? Good bye torrenting.