Based on that description I'd say it's closer to taking a RAW format and converting it into a JPG, which contains most of the (human visual) relevant data from a DCT while removing the remaining "noise"... which is a bit different from simply resizing (eg subsampling). What I would find a bit satisfying about this description is that you probably wouldn't get away with claiming that a JPG of an image was un-copyrighted and only a BMP (of the same resolution) was.

To be fair, it does depend on the amount of compression before it is not recognizable, but if you can still squint and see the Mona Lisa (when you also have her phone #)... have you not violated her privacy?