Evolution can both add and remove junk DNA. Humans are descended from bacteria.

More particularly, the equilibrium size of the DNA is very roughly inversely correlated with population size. A larger population size is better at filtering out disadvantagous traits. It's not linear - there are discontinuities as decreasing population size eliminates natural selection's ability to select against different things. And those things sometimes can even go on to be selected for for other reasons - there are genomic structures that are important for eukaryotes that could probably never have evolved in a bacterium because to get to them you need to go through various local minima of fitness.

Soil bacteria can have trillions of individuals per cubic meter of dirt and they actually experience direct evolution towards lower genome size - more DNA means more sites at which something could mutate and become problematic and they actually feel this force. Eukaryotes go up in volume by a factor of ~1000 and go down in population by at least as much, and lose much of the ability to select against introns and middling amounts of intergenic DNA and expanding repeat-based centromere elements.

Multicellular creatures with piddlingly tiny population sizes compared to microbes lose much of the ability to select against selfish transposon DNA elements, gigantic introns and gene deserts, and their promoter elements get fragmented into pieces strewn across many kilobases rather than one compact transcriptional regulation element of a few dozen to a few hundred base pairs (granted, we've also been able to make good use of some of these things for interesting purposes from our adaptive immune system to the concerted regulation of our hox gene clusters that regulate our body plans). They also become very sensitive to the particular character of the transposons or DNA repair machinery of their particular lineage and wind up random-walking like crazy up and down an order of magnitude or two in genome size as a result.