When someone uses AI against you

The guides on this site are about you choosing to use AI. This is about someone else using it, on you, without asking. The good news is that the defense is simple, low-tech, and takes one phone call to set up.

The call

The phone rings, late. It's your grandson, or your daughter, or your brother. The voice is unmistakably theirs and they are frightened — there's been an accident, or an arrest, and they need money now, and please don't tell anyone else in the family.

It isn't them. The voice was assembled by software from a few seconds of audio, and the person on the phone has never met your family.

Why the advice you were given stopped working

For twenty years, the warnings about scams were about tells. Look for bad grammar. Listen for a voice that doesn't sound quite right. Watch for the generic greeting that doesn't use your name.

Every one of those tells is gone. The message is now written flawlessly. The voice sounds exactly right because it was built from your relative's actual voice, taken from a birthday video, a podcast appearance, a voicemail, or any of the other places a few seconds of someone talking sit in public. Advice built on spotting a flaw doesn't survive tools that don't make flaws.

So the defense has to move. It can no longer be can I tell this is fake? It has to become can I confirm this is real? — a completely different question, and one you can still answer reliably.

Set this up tonight. It takes one phone call.

Agree on a word with the people who might call you in an emergency. Any word, as long as it's odd enough that it wouldn't come up naturally and has never appeared anywhere online. Everyone memorizes it; nobody writes it in a text message or an email. If a distressing call comes, you ask for the word before anything else. The voice can be copied perfectly. The word cannot be guessed.

The shape of the call, so you recognize it

The specifics vary but the structure almost never does. Knowing the structure is most of the protection, because recognizing a script in progress breaks the spell it depends on.

Urgency. Something terrible has happened and there are minutes, not hours, to fix it. Real emergencies rarely come with a deadline that tight.
Secrecy. "Don't tell Mom." "Please don't call Dad, he'll be furious." This exists purely to keep you from making the one call that would end the scam, and it's the single most reliable warning sign there is.
A second voice with authority. A lawyer, an officer, a doctor takes the phone to confirm the story. Two voices agreeing feels like corroboration. It's the same person.
A payment method you can't reverse. Gift cards, a wire transfer, cryptocurrency, a courier coming to your door for cash. No court, hospital, or police department has ever asked for bail money in gift cards.

The two habits that stop it

Hang up and call back on a number you already have. Not the number that called you, and not a number the caller gives you — both can be faked, cheaply and convincingly. Use the contact already in your phone. If they answer and they're fine, you have your answer. This one habit stops the overwhelming majority of these calls, and the few seconds of awkwardness if you're wrong cost you nothing at all.
Ask for the word. If you set one up, this is instant and certain. A real relative in real trouble will be confused for a second and then tell you. Nobody else can.

Both habits work because they don't require you to judge whether the voice sounds right. They sidestep the question entirely, which is the whole point — under that much stress, nobody's judgment is good, and the scam is designed around exactly that.

The same trick, other doorways

The phone call is the version that costs people the most, but the same technology shows up elsewhere: emails written in a colleague's exact style, video calls where the face moves convincingly, months-long friendships or romances built by someone who never types a word themselves, and investment pitches featuring a well-known person who never said any of it.

The response is the same in every case. Verify through a channel the other person didn't choose. Call the number you already had. Walk down the hall. Ask something only they would know. Anything that steps outside the conversation they control.

If it already happened

Call your bank first — some transfers can be stopped if you move quickly, and the first hour matters more than anything else you'll do. Then report it, to your local police and to the FTC at reportfraud.ftc.gov, or the FBI's complaint center at ic3.gov.

And please don't carry this quietly. These scams are underreported mostly because people feel foolish, which is precisely the outcome the scammer is counting on. Being fooled by a perfect recreation of your grandchild's frightened voice is not a failure of intelligence or attention. It's a failure of a defense that stopped working, and you now have a better one.

Why this is on this site

Because this is what the fear is actually about. When people say AI frightens them, this is usually closer to what they mean than anything about chatbots or jobs, and it's a reasonable thing to be frightened of.

But fear alone leaves you at the mercy of the next convincing phone call. Understanding what the technology can and cannot do is what turns that call from a catastrophe into an inconvenience — you ask for a word, you don't get it, you hang up, you call your grandson, he's asleep. That's the whole difference, and it's available to anyone who reads this page.