A malware article in CSO Magazine points out the ultimate futility of the signature-based antivirus detection and blacklisting mechanisms, given the escalating rate of release of new/variant malware and its inability to block data theft (which is what Data Leak Prevention is all about: personally, I never expected AV software to do this so that is a rather curious point).
The demise of signature-based AV detection has been predicted many times before but it stubbornly remains a relatively effective and inexpensive control, on the whole. I'm worried about bespoke malware, custom-written to infiltrate specific target organizations, but there other techniques come into play, DLP and checksumming being two of them. So called "heuristic scanning" has a bad press for generating too many false positives, but that's another piece of the defense-in-depth puzzle, along with prompt patching and (of course) security awareness. There's no need to detect avoided malware.