Synthetic Voices Raise Fresh Questions About Consent

Thirty seconds of recorded speech is now enough to build a working copy of a voice — cheerful, tired, persuasive, young again. The tools are startlingly good, cheap to run and already woven into audiobooks, games and customer helplines. Regulators, performers and families are all asking the same question: who gets to say yes?

Consent is the bright line most experts draw. A narrator licensing her voice for a series is a transaction; a stranger cloning a neighbour's voice for a prank call is something else entirely. Between those poles sits a grey field of fan projects, memorial voices and political satire.

Where harm appears first

Fraud teams report the sharpest rise in voice-based scams, where a familiar voice pleading for help can bypass scepticism in seconds. Banks and phone networks are responding with verbal passwords and call-back rules — and with a simpler message: hang up, then call back on a number you trust.

"A voice used to be proof of presence. It is now proof of nothing — and our habits have not caught up with that fact."

Performers' unions, meanwhile, are negotiating likeness clauses that treat a voice the way image rights treat a face: licensable, revocable and payable. Several studios have adopted standard contracts this year.

A practical path forward

Ethicists propose three defaults: watermark every synthetic voice, require explicit consent for identifiable copies, and ban real-time impersonation in calls outright. Technology to detect clones is improving — but experts warn detection will always lag behind generation, making norms and law the real guardrails.