When AI Sounds Like Family — Bryan "B" Kemler

The phone rings just before midnight. The voice on the other end is your grandson’s. He’s crying. He’s in another state, or another country, and the story comes out fast — a car accident, a hospital, a police officer who wants to speak with you, an attorney who needs a payment to keep this quiet so it doesn’t go on his record. Please don’t tell his parents. He’ll explain everything in the morning.

You know that voice. You have known it since he was four. Every cadence, every nervous laugh, the way he says Grandma when he’s scared — all of it is exactly right.

Except it isn’t him.

It isn’t even a person. A computer generated the call from a thirty-second clip of his voice — pulled, most likely, from a high school graduation video posted publicly on a relative’s Facebook page two summers ago. The script is being typed in real time by a stranger several time zones away. The money, if you wire it, will be in three different countries by morning.

This is the most common version of what’s now happening to American families every day. The Federal Trade Commission has tracked impersonation scams across hundreds of thousands of complaints and well into nine figures of reported losses, and the AI-voice subset is the fastest-growing slice. It is no longer a fringe story. It is the new shape of an old crime.

I want to give you a single, free, ten-minute defense against it.

What’s actually changed

Until about three years ago, a convincing voice impersonation took a skilled human performer and a quiet recording studio. It was expensive, slow, and rare. Today the same result takes a phone, a thirty-second audio sample, and a piece of software that costs less per month than a streaming service. The audio sample doesn’t need to be a confession or a phone call. A wedding toast does it. A church reading does it. A podcast interview, a coaching video, a TikTok, a voicemail greeting — any of them does it.

The pool of available audio for almost any American under forty is enormous. The pool for many Americans over forty is also large. If your son is a pastor, your daughter is a teacher, your grandson plays in a band, your husband narrated a community history project — there is audio of them, somewhere, that a stranger can find in under an hour.

Once a scammer has thirty seconds of someone’s voice, they can have that voice say anything. The technology does not know whether the sentence is true or false. It does not care that the person is asleep in their own bed, two thousand miles away. It will say “Grandma, please don’t be mad at me” in your grandson’s exact voice, on demand, at 11:47 p.m.

The familiar voice has always been the password. That’s why this crime works. The voice is no longer trustworthy as identification — not because anyone wanted that to be true, but because the technology has changed and the law and the culture have not yet caught up.

What still works

The countermeasure is older than the threat, which is part of why it works. It is something families used to do as a matter of course, long before the internet — a verbal code, agreed on in person, that the family uses to verify each other in unusual situations.

I call it a Verified Family Code.

A Verified Family Code is a short phrase, two or three words, that you and the people you love agree on together — out loud, in the same room — and never write down, never email, never text, and never say on a recorded call. It can be a word from a shared family story. It can be a nickname for the dog from twenty years ago. It can be a phrase from a song the children sang on a road trip. The only requirements are that it be memorable to your family and unfindable by anyone outside it.

When a phone call comes in claiming to be a family member in trouble, the rule is simple: I can’t help you until you tell me the code.

The reason this works against voice-clone scams is that the technology that synthesizes your grandson’s voice can only say what someone tells it to say. The scammer typing on the other end has access to everything public about your family — names, schools, jobs, recent vacations, even, sometimes, the names of pets. They do not have access to a phrase that has never been written, posted, recorded, or transmitted electronically. They cannot guess it. They cannot Google it. They cannot get it out of a chatbot. They cannot fake it without being told what it is, and there is no way for them to be told.

When the voice on the phone cannot produce the code, you have your answer, and you can hang up.

It is worth saying clearly: the real grandson, asked for the code in a real emergency, will be momentarily confused, then relieved you remembered to ask, and then he will say the words.

How to set one up tonight

You don’t need a lawyer for this. You don’t need software. You don’t need to buy anything. The four steps fit on an index card.

The first step is to call a family meeting — in person, at the next dinner, at the next holiday, or simply when the family is together for any reason. Phone is acceptable if in-person is not possible, but only on a normal phone call, not a recorded one, and not in a place where smart speakers can hear.

The second step is to choose the phrase together. Two or three words. Memorable to all of you. Connected to a shared memory. Not the name of a pet, child, school, or street that appears anywhere online or in your public records.

The third step is to agree on the rule. Any phone call that asks for money or urgent action — bail, a ransom, a medical emergency, a wire transfer to keep a problem quiet — requires the code before anyone does anything. Even if the voice sounds exactly right. Especially if the voice sounds exactly right.

The fourth step is to commit it to memory. Do not write it down. Do not put it in a password manager. Do not text it. Do not say it on a recorded line. The only place it lives is in the heads of the family members who agreed to it.

If your family is large or multigenerational, you can layer the code — one phrase for the immediate family, a different phrase for the broader circle, refreshed every few years when everyone is together.

Where this fits

A Verified Family Code is a single, free tool. It is also part of a broader category of protection I think every family — and especially every family with significant assets and elderly members — should consider in the AI era. A periodic check of what the public internet already knows about your family. A digital inventory the people who’ll handle your affairs can actually find. A clear set of instructions for the trusted people who will step in if a parent’s judgment slips faster than the family expects.

None of these are new ideas. What’s new is the threat profile, and the small set of protective practices a family should add to what they were already doing. This is the work I do through LegacyGuard: calm, concrete protection for families, without the scare tactics. Most of it is closer to common sense than to computer science.

Set up your code this week. It costs nothing, it takes ten minutes, and it’s the rare security measure that gets stronger the simpler you keep it. The next time that midnight call comes — and for a lot of families, it eventually does — you’ll have an answer ready that no machine can fake.