Voice Cloning: Verify the Request, Not the Familiar Voice
A convincing voice is not proof of identity. Build a separate verification path for payment changes, urgent requests and account recovery.

A familiar-sounding caller can create a powerful sense of urgency. Voice cloning makes familiarity a weaker identity signal, but the defensive response does not require employees to become audio-forensics experts. It requires a process that verifies the request through a channel the caller does not control.
Break the caller's control of the conversation
The US Federal Trade Commission has warned that scammers can use voice cloning to imitate people in distress and recommends contacting the person through a number already known to be theirs. For business workflows, the same principle supports a callback using an independently maintained directory, not a number supplied during the suspicious call.
A callback is one layer, not a complete guarantee. Compromised accounts and phone systems can undermine a second channel. Sensitive actions should also follow existing authorization and approval rules, especially when bank details or recovery methods are changing.
Prepare the workflow before the urgent call
- Define which requests require an independent callback or written approval.
- Use contact details from a trusted directory rather than the incoming message.
- Require a second authorized approver for high-risk payment changes.
- Give employees permission to pause an urgent request without penalty.
- Train on pressure tactics and escalation paths rather than on spotting audio glitches alone.
A realistic exercise presents an urgent but unverified request and measures whether the employee follows the process. Avoid relying on a static secret phrase as the only safeguard: it can be disclosed, recorded or reused. The goal is a verifiable decision path even when the voice sounds entirely convincing.
Sources and scope
Prepared analysis applying the FTC's 2023 voice-cloning warning to business verification procedures. It does not report a future fraud incident or claim that audio detection is foolproof.

