Yes it seems to be in some sense the reverse of somatic hypermutation. Rather than have particular regions targeted for high rates of mutation by enzymes that make mutations occur, some genomic regions are instead much more protected against mutation by different protective mechanisms.
Of course the finding that mutations are not distributed with equal probability across the genome as a whole, and that there are mutational hot spots and cold spots, has been known for a long time now.
Does a bias in a probability distribution make the process non-random? Mutations still appear to occur everywhere, just with unequal frequency, so in some places more than others. And in places with critical and house-keeping genes they’re more protected.
That doesn’t seem to me to substantiate any claim that mutations are directed to be more adaptive(as if the organism knows what kinds of mutations will be adaptive), since the organism can’t know whether mutations in a house-keeping gene might turn out to be adaptive in some future circumstance. This biasing of mutations to be lower in some regions is analogous to an adaptive, but ultimately still probabilistic behavioral strategy. Act in a certain way because this has in the past, on average, been less likely to be detrimental.
So no, I don’t think this is actually “challenging the prevailing paradigm that mutation is a directionless force in evolution.” It looks like textbook evolutionary game-theory, but with mutation biases instead of behavior.