Social Engineering

Generative AI lowers the cost of social engineering by orders of magnitude. Spear-phishing emails that previously required a fluent writer and target research are now produced in seconds with reasonable per-target personalisation. Voice cloning (ElevenLabs, OpenVoice, and others) enables real-time impersonation of executives and family members; multiple confirmed business-email-compromise and CFO-fraud incidents in 2023-2024 used cloned voices. Deepfake video is good enough for short verification clips and live calls under poor video conditions. Beyond direct attacks, AI-generated content fuels disinformation campaigns, fake review economies, and pig-butchering scams at unprecedented scale. AI Threat Alert tracks this category through CVEs in voice/face-recognition systems that fail to detect synthetic media, plus incidents in AIID (the AI Incident Database). Defenses: out-of-band verification for sensitive actions, deepfake detection layered with provenance signals (C2PA), and user education that assumes any voice or video can be faked.