I believe Dropbox uses a method analogous to block-level dedupe. That is, files are split up into smallish chunks and then the chunks are what get "deduplicated". A "file" basically consists of a list of pointers to chunks.
This makes things extra problematic because completely unrelated files might share chunks. Standard file formats may lead to duplicate headers. Or consider a political science textbook that contains a complete copy of the US Constitution, and a file that contains just the US Constitution. One is perfectly legal to distribute freely, the other may not be, but both might share some common blocks, and a federal judge with a shoot-first mentality might craft an order requiring the deletion of those common blocks.
This makes things extra problematic because completely unrelated files might share chunks. Standard file formats may lead to duplicate headers. Or consider a political science textbook that contains a complete copy of the US Constitution, and a file that contains just the US Constitution. One is perfectly legal to distribute freely, the other may not be, but both might share some common blocks, and a federal judge with a shoot-first mentality might craft an order requiring the deletion of those common blocks.