Hey guys, Non here. Coffee in hand.
The voice you are hearing is a copy of mine. I trained it on recordings of me. I am using it on purpose.
This week, more than eighty British performers wrote to the Prime Minister. They want the law to say every person in the United Kingdom owns their own voice.
They are not wrong. They are also not the whole story.
I want to walk you through who this law would actually protect, and who it would leave hanging. Because I am sitting on both sides of it.
The letter went to Prime Minister Andy Burnham. The campaign is called Save Our Voices Now. Among the names on it: the actor Hugh Bonneville, from Paddington and Downton Abbey. The actor Matt Lucas. The actress Nicola Coughlan.
Hugh Bonneville put the claim in one line. The Guardian printed it. "My voice is unique, it is mine, it is part of my identity. So is yours."
He added a note that the quote was handwritten, not created by AI. That is the joke, and it is also the fear.
The campaign was started by two people who live on this problem every day. One is an audiobook narrator, Alice Sockett. The other is a performer, Peter Caulfield.
Alice Sockett told the BBC her voice is already out in the public domain, and completely unprotected. She called cloning an existential threat to the whole industry. Peter Caulfield told the Guardian that a voice can be copied from a few seconds of audio. With the right system, that copy can then be made to say whatever the prompt wants.
The BBC's own reporting is more careful on the time. Quick clones can be made in minutes from a few seconds of audio. High-quality copies still tend to need hours of training. Both numbers matter. Minutes is the scam. Hours is the replacement of a working narrator.
Here is the incentive, and it is not a mystery. If a company wants an audiobook or a commercial, they used to hire a person to sit in a booth. Now they can train a model on recordings that already exist, and generate infinite variations of that person's sound.
A one-time "voice skin" is cheaper than a human being, every time, on every new job. People are driven by incentives, not by speeches about the soul of art. Follow the invoice and the story explains itself.
An actress from Derry Girls told the BBC audiences want human connection, not AI. A Scottish singer, Sandi Thom, said her voice was handed to her at birth, with the life and the pain behind it.
Fair enough. A voice is not a spreadsheet. Anyone who has heard their own mother on the phone already knows that.
Now, the version you will see in the headlines is that this is a battle for the soul of human creativity. Human versus machine. Artists versus theft.
That framing is very effective. It is also a flatline. It has a beginning and an end and gives you no reason to keep thinking.
So let me give you the part that actually has a crack in it.
The same week, the BBC aired the other load this technology is carrying. Yvonne Johnson lost her ability to speak because of motor neurone disease. She had her voice cloned. She told the BBC's Tech Life programme it brought her closer to her family, who had started to forget what she sounded like.
In that room, the clone is not a replacement for a person. It is a stand-in voice for a person biology is trying to silence.
There is a second case. I will do it once, carefully. Pam Cronrath, in Washington state, used the same class of tool at her husband's funeral. A digital version of him addressed the people who came. She told BBC News she was completely impressed. It stayed with her.
You can hate that. You can also see why a widow would want it. A good law has to survive both rooms, not just the one with famous names on the letter.
The campaign says twenty-eight percent of UK adults report being targeted by a voice-cloning scam. That number is theirs. I am repeating it as their claim, not as a census.
Peter Caulfield's point underneath it is the one that lands. If your bank, your parent, or your boss hears your voice asking for money, the question will not be whether the audio sounded convincing. The question will be whether the law was written before the call came in.
Denmark is already moving. The Guardian reports reforms there that give people legal ownership of their face, their body, and their voice. They can demand that unauthorised AI copies come down, and they can seek compensation.
Britain is offering a consultation. A government spokesperson told the BBC that digital replicas can be a powerful tool, including for the creative industries, and that they can also do harm. They say they will consult on how to address the harms while protecting "legitimate innovation."
Consultation is what you do when you have not decided who the tool is for.
A lawyer who studies this told the BBC in June that the law comes close. It does not fully protect your voice, your face, or your identity. Unfit for purpose is the phrase that has been hanging over this for months. It is not a slogan. It is a missing category.
Here is the building, read like an engineer.
The load is cheap, infinite speech, trained on recordings people already gave away. The crack is that British law still does not know whether a voice is property you can own, or information anyone can use. Who signed off on the gap? A government that will consult, while the invoice has already moved.
The cost of doing it right is not a ban. A ban would protect the actors and hang Yvonne Johnson. The cost of doing it right is a rule with three parts.
One: you cannot take my voice without my name on the paper. Two: if you use it to make money, I get paid. Three: a person who has lost speech can still build a speaking copy of themselves.
That is consent, payment, and a medical carve-out. It is not poetry. It is a spec.
Someone will say I would say that. I run a daily show on a clone of my own voice. Of course I want the tool to stay legal.
Yes. I would. I also wrote the recordings. I know what went into this copy. Nobody scraped me in a parking lot. That difference is the whole law.
If you cannot tell those two cases apart, you will write a rule that punches the wrong person.
What should you do with this tomorrow morning?
If you make things with your voice, start with the contract. Put the ownership sentence in before the session starts. If you build these tools, stop hiding the training set. If you are just a listener, the next time a family member calls asking for money, call them back on a number you already have.
And if you are listening to me: this voice is mine because I said so, on the record, in public. That is the standard I would hold everyone else to.
A voice is not a dataset. It is also not sacred. It is a tool with a name on it.
Put the name on the paper. Then we can talk.