As Nvidia had fairly recently learned, blockades from running certain algorithms will be circumvented. Many applications also use GPGPU for acceleration of non-graphics applications (GPUs are pretty much highly parallel supercomputers on a chip), so cutting off GPGPU is not on the table either. Unless you wish to just completely screw over the open source community, and go the white list route.