
Audio and Video Deepfakes Are the New Social Engineering Weapon
Imagine a colleague calls you and asks you to urgently approve a wire transfer before the end of the day. The voice sounds exactly right. But…
Do you pause to double-check? Or do you process the request because you recognize the voice?
Recent research shows that 41% of CISOs reported at least one social engineering incident involving an audio deepfake over the last 12 months. On top of that, 36% of security leaders reported employees being targeted by deepfake video calls during the exact same period.
The tools generating these realistic voice clones are cheap, fast, and accessible to anyone. A bad actor needs only a few seconds of publicly available audio from a podcast, panel discussion, or social clip to clone a voice. Once synthetic media bypasses human instincts, traditional security filters downstream become irrelevant.
Deepfake Attacks Target Trust: Traditional IT Security Is Not Enough
AI is dramatically increasing the volume, personalization, and credibility of social engineering attacks while stripping away the obvious red flags employees used to rely on.
Hearing a familiar voice triggers instant trust before critical thinking even kicks in. Criminals exploit this confidence to target high-stakes operations like payment authorisations, Account recovery, and even privileged access requests.

Audio and video deepfakes are not replacing traditional email phishing. They are layering directly on top of them. A target who gets a legitimate-looking email setting up context is much more likely to approve a follow-up voice request without second-guessing the caller.
Technical barriers like managed firewalls, endpoint detection, and email gateways are essential for technical security. However, technical systems alone were not built to verify human identity during a live phone call or video chat.
Even regulatory frameworks point to this blind spot. Guidelines like the Monetary Authority of Singapore (MAS) Technology Risk Management expectations require strict controls around authentication and access. When a deepfake tricks an employee into granting unverified access, it creates a break in those compliance controls even if all server firewalls are running perfectly.
Social Engineering Gap Where Managed IT Security and Email Filters Were Not Built For
The solution is shifting toward default verification protocols.
Default verification does not ask employees to guess whether an audio file or video call is real. Instead, it enforces a strict process for every high-risk request, regardless of who appears to be asking:
- Out-of-band callbacks: Confirming requests by hanging up and calling back a pre-verified number stored in official company records.
- Secondary channel approval: Requiring confirmation through dedicated internal tools before initiating wire transfers.
- Mandatory cooling-off pauses: Enforcing a pause on high-risk credential or payment changes.
Security awareness training that embeds this kind of verification behaviour rather than training people to spot technical tells is what addresses the deepfake threat.
