Field Notes — N° 01
voice/vɔɪs/
n. — L. vox, that which is uttered; a sound carrying intent, identity, presence.
English
voice
Español
voz
中文
声音
हिन्दी
आवाज़
The Specimen 0:00  /  0:00
Recorded by Jeff McMillan. Cloned voice rendered in English · Español · 中文 · हिन्दी.

For the past few weeks I have been working to perfect my AI cloned voice.

Fifteen minutes to create something mediocre. A week to get it to a reasonable spot.

A training module that took me an hour to record now takes ten minutes. And I can personalize at scale — by industry, by role, by country. Want my program in Japanese? A day. Customized for healthcare professionals? Not a problem.

And the positive applications go far beyond business.

A stroke victim can be given their voice back. A grandmother can read bedtime stories to grandchildren born after she is gone. The oral histories of indigenous tribes can be preserved in the original voices.

This is one of the most humane technologies we have ever built. It is also one of the most dangerous.

The elderly are getting calls from "family" asking for money via Venmo and Zelle.

Executives are wiring millions based on a call from "the CEO."

Parents are getting calls from a kidnapped "child" — sobbing, terrified, demanding ransom. The child is in school.

As election day approaches, fabricated audio of candidates will land in voters' inboxes hours before polls open. The retractions will arrive after the votes are cast.

We can no longer trust what we hear.


Three things need to happen — and we are already late.

i.

Authentication has to be rebuilt.

Any institution that relies on voice as proof — banks, brokerages, executive offices, government — must move to multi-factor verification. Voice alone is no longer a signal.

ii.

Families need codes.

Pick a word. Share it with your spouse, your children, your parents. If "family" calls in distress, the code gets asked. No code, no money. Every adult should adopt this habit this week.

iii.

Policy has to catch up.

Mandatory disclosure of AI-generated audio. Criminal penalties for malicious cloning. Cryptographic provenance standards. The technology will not pause. The law must move.

The same tool that lets me record a training program in any language from my desk lets a stranger drain your mother's account using your grandson's voice.

The harder work isn't building this.
It's deciding what we will not tolerate now that it exists. We need to start.

Want to talk about what this means for your organization?

McMillanAI advises boards, CEOs, and leadership teams on navigating exactly this kind of moment.

Start a Conversation