Good point. Future versions of the project could theoretically work with encrypted blobs, with the keys (which are significantly easier to keep in multiple physical locations) only known to the Internet Archive. This would probably require some architectural changes to the Archive and might be harder to scale, plus one of the goals of IA.BAK is to not be dependent on any Internet Archive resources other than the actual files.

Definitely something to think about down the road, but in the meantime there are plenty of public datasets to back up.