This is not an in-depth post, just a very basic introduction to the carnivorous sponges, in response to a reader question.
Before anything, yes, carnivorous sponges do exist, but they are taxonomically very limited: only some members of the Poecilosclerida order of demosponges are known to be carnivorous, and there, most belong to the Cladorhizidae family, with a couple of guitarrid and esperiopsid species as well.
They have convergently developed a range of morphological similarities: their symmetrical body is held on a stalk. From the body, filaments stick out laterally, and they are covered in hook-shaped spicules (anisochelae). Except in one genus of carnivorous sponges, Chondrocladia, carnivorous sponges have lost two of the autapomorphies of the sponges, the aquiferous system and the chambers where the choanocytes are housed.
That these modifications have been independently reached by most carnivorous sponges is a clear hint that they are functional morphological adaptations. The latter two effectively rule out the normal filter-feeding way of life of a sponge. The stalk allows the sponge to grow to higher heights, where its prey, small crustaceans, are likely to be swimming. Once a prey item lands on one of the filaments, it cannot detach itself – the spicules act like “Velcro”, as Vacelet & Boury-Esnault (1995) describe. Massive cell migrations then occur towards the epithelium of the filament; eventually, the prey will be swallowed whole and presumably digested.
They are mostly found in the deep sea, where the lack of nutrient-rich currents is a clear selection pressure that can be imagined as leading a filter-feeder to adopt a carnivorous habit.
Vacelet, J., & Boury-Esnault, N. (1995). Carnivorous sponges Nature, 373 (6512), 333-335 DOI: 10.1038/373333a0