Deepfakes: Vulnerability and risk in malicious use of advanced AI