The usual p-type dopants of SiC do not provide sufficiently shallow acceptor levels. It is argued that the reason for that lies in the large relaxations around substitutional B and Al. The complex AlSiNCAlSi is suggested as an alternative. First principles calculations show that the lattice relaxation around this defect is smaller, and that it gives rise to an acceptor state more delocalized and shallower than that of aluminum. The complex is predicted to be stable with respect to its constituents.