The pdfs don't have arbitrary content, they contain very specific headers (crafted by Google research), in one file the header selects one image, in the other i

The pdfs don't have arbitrary content, they contain very specific headers (crafted by Google research), in one file the header selects one image, in the other it selects the other. These two headers are SHA1 colliding and the remainder of the files are identical.