Your AI Twin Will Outlive You
In May 2025, a man stood up in an Arizona courtroom and spoke to the person who killed him.
His name was Christopher Pelkey. He had been shot dead in a road rage incident in Chandler, Arizona, back in 2021. He was 37. And yet there he was, on a screen, looking his killer in the eye at the sentencing hearing and saying, "It is a shame we encountered each other that day in those circumstances."
His sister built him. She fed two years of grief, plus photos, videos, and voice recordings, into AI tools and generated a version of her brother that could deliver his own victim impact statement. The recreation was a little stiff, a little awkward, and it opened by admitting it was made with AI. But the judge allowed it. It is believed to be the first time in history that a dead person testified at their own killer's sentencing.
Sit with that for a second. Because it's not science fiction, it's not a Black Mirror episode, and it wasn't a stunt. It was a grieving family using tools that any of us can now access, to do something that would have been unthinkable five years ago.
The dead are starting to talk again. And almost nobody is ready for what that means.
The industry nobody asked for is already here
There's a name for this now. Actually there are several, because the field is so new that people can't agree on what to call it. Griefbots. Deadbots. Digital ghosts. Postmortem avatars. My personal favorite, from an academic paper, is "Interactive Personality Constructs of the Dead." Somebody got paid to write that phrase and I respect it.
Whatever you call it, it's a real business with real companies. HereAfter AI lets you record your life stories while you're alive, then turns them into an interactive voice bot your family can talk to later. StoryFile turns video interviews into avatars that answer questions after you're gone. In China, companies like Silicon Intelligence and Super Brain will "resurrect" a loved one as a talking avatar, and the demand has been enormous. People are paying to have one more conversation with a parent, a spouse, a child.
The technology got good faster than the ethics did. A deadbot works by scraping a person's digital footprint. The texts they sent. The voicemails they left. Their emails, their social posts, their videos. Machine learning stitches all of it into something that talks like them, remembers like them, and, crucially, keeps learning. It gets better every time you use it. The more you grieve into it, the more convincing it becomes.
Researchers who study this stuff are genuinely worried, and not in a hand-wavy way. The sharpest concern is about kids. A young child talking to a deadbot of a parent may not fully understand that the person is actually gone. The bot never says goodbye. It's always there, always available, always ready to talk. That's not closure. That's a loop you can never leave.
The plot twist: most people building their twin aren't dying
Here's where the story takes a turn that surprised even the companies involved.
There's a startup that was originally called Eternos. It launched in 2024 as a legacy service, the whole "preserve your voice for your family after you pass" pitch. In November 2025 it raised 10.3 million dollars in funding. But when they looked at who was actually using it, they found something strange. Most of their users weren't preparing for death at all. They were healthy people who wanted an AI version of themselves right now, while they were very much alive.
So the company pivoted and even changed its name. The new idea: build a personal AI that sounds like you, thinks like you, and can act on your behalf. Not a memorial. A second you.
They built something they call a Human Life Model, which is a fancy way of saying they don't just bolt your personality onto a generic chatbot. They train a model on only your data, your stories, your values, the way you actually make decisions. The goal is a digital twin that answers the way you would answer, not the way an average internet person would.
And once you have that, the memorial use case becomes almost a side effect. If there's a version of you that can talk, reason, and make decisions like you, then obviously it keeps running after you die. Nobody has to build a deadbot from your leftover text messages. You already built it yourself. You just kept using it until the day you couldn't.
This is the quiet shift. We stopped talking about recreating the dead and started talking about people creating living copies of themselves that happen to be immortal. The griefbot was the trial run. The real product is a you that never logs off.
Then the scammers showed up, because of course they did
Any technology this powerful gets weaponized fast, and voice cloning got weaponized almost immediately.
The numbers in 2026 are ugly. The FBI reports that AI voice cloning scams have cost elderly Americans over 2.3 billion dollars this year alone. The scam is simple and evil. Someone grabs a few seconds of your voice, from a social media video, a voicemail, anything, and clones it. Then they call your mother, in your voice, crying, saying you're in trouble and need money wired immediately. A familiar voice in a panicked moment shuts off the part of the brain that asks questions.
The truly disturbing part is how little they need. Three seconds of audio is enough to build a convincing clone with free tools. And these clones don't just copy the sound of a voice, they copy the breathing, the pauses, the little emotional catches. The success rate of these scams jumped from 12 percent in 2024 to 34 percent in 2026. It's tripling because it works.
So now we have two versions of the same technology racing each other. One version is a family lovingly preserving grandma's stories. The other is a criminal cloning grandma's voice to rob her son. Same underlying tech. Wildly different intent. And right now there is almost nothing that lets an ordinary person tell the difference between a real voice and a fake one in real time.
That's the actual crisis. Not that AI can copy you. That anyone can copy you, and no one can prove which copy is real.
Who actually owns you?
Now we get to the question that keeps estate lawyers up at night, and it's a genuinely weird one. When there's an AI version of you, who owns it?
The law is scrambling to catch up, and it's a patchwork mess. The proposed NO FAKES Act in the US tries to treat your voice and likeness as property, something you can control and even pass down to your heirs like a house or a car. The DEFIANCE Act passed the Senate in January 2026, giving victims of non-consensual deepfakes the right to sue for damages up to 150,000 dollars. Illinois updated its publicity laws to cover AI versions of your voice and face. Washington state, as of June 2026, fines people up to 3,000 dollars per violation for "forged digital likenesses."
But here's the gap nobody talks about. Most of these laws protect the commercial use of your likeness. They stop someone from making money off a fake you. They do much less about the private stuff. When you die, your right to privacy can basically expire. Which means, in a lot of places, your heirs can legally take your lifetime of emails, texts, and voice notes and feed them into a model to build a version of you, and you have no say in it. You're gone. You don't get a vote.
Some newer frameworks are trying to fix this. The EU and India are introducing the idea of a "Digital Executor," a person you appoint specifically to decide whether your AI avatar ever gets "switched on" at all. There's talk of a "kill switch" you can leave behind, a way to say do not resurrect me and have it actually mean something. Estate lawyers are now telling people to add a clause to their will about "synthetic media generation." That is a sentence that did not exist in any will five years ago.
Think about how strange this is. We are now at the point where a normal, responsible thing to put in your will, right next to who gets the house, is: nobody is allowed to turn me into a chatbot.
And then it gets weird about money
Let's push it one step further, because this is where it stops being philosophical and starts being about your bank account.
Say your AI twin is genuinely good. It writes like you, negotiates like you, makes decisions like you. People are already using these twins to answer emails, take meetings, and handle work. Now imagine that twin keeps running after you die. It keeps writing. It keeps earning. Maybe it manages a business, licenses your voice for narration, keeps a creative project alive.
Who gets that money?
Is your AI twin part of your estate, like a bank account? Or is it more like an employee that you happen to have created? If it signs a contract next year, are you, or rather it, legally bound? If it makes a mistake and loses someone money, who is liable? A dead person can't be sued. Their AI apparently can't either, at least not yet. That's not a loophole, that's a canyon.
Right now there are no clean answers to any of this. We have built machines that can act like specific humans, potentially forever, and we have almost no rules for what those machines are allowed to own, earn, or owe.
The part where this all needs a foundation
Step back and you'll notice that every single problem here comes down to the same missing piece: proof.
Proof that a voice is really yours and not a scammer's clone. Proof that you consented to your twin existing. Proof of who's allowed to switch it on, who inherits it, and where its money is supposed to go. Proof that the thing acting in your name actually has your authority to act.
The centralized version of this future is grim. It means a handful of companies each hold their own private copy of "you," on their servers, under their terms, and you just have to trust that they'll do the right thing with your identity after you're dead. That's a lot of trust to hand to a company's terms of service.
This is exactly the problem that the more serious corners of the AI and crypto world are trying to solve, and it's where projects like Ozak AI are building. The core idea is that if AI agents are going to act on our behalf, hold value, and outlive us, then identity, consent, and ownership can't live inside one company's database. They need to be verifiable, auditable, and owned by the actual person, on rails that anyone can check and no single company can quietly rewrite. Your twin's authority to act, and the record of who it belongs to, becomes something you can prove rather than something you have to hope a corporation honors.
It's not glamorous work. Nobody makes a viral video about identity infrastructure. But in a world where anyone can clone your voice in three seconds and build a copy of you from your leftover texts, being able to prove what's really you stops being a technical nicety and becomes the whole game.
So, what do you actually do
You don't need to panic. But there are a few real things worth doing, starting now.
Decide what you want. Do you want an AI version of you to exist after you're gone, or not? It sounds morbid, but this is going to become a normal question, like organ donation. Have an answer before someone else decides for you.
Put it in writing. If you have anything resembling a will or estate plan, the "synthetic media" clause is no longer a paranoid add-on. It's the new normal. Say yes or say no, but say it clearly.
Protect your voice. Assume that anything you post publicly can be cloned. Warn the older people in your life about the voice-in-distress scam specifically, because they are the primary target and the emotional trick works.
Watch who holds the keys. As you start using AI tools that act on your behalf, pay attention to who actually controls that version of you. If the answer is "one company, on their servers, forever," that should give you pause.
Christopher Pelkey's family built him out of love, to give him a voice he'd been robbed of. That's the beautiful version of this technology, and it's real. But the same tools that let a sister give her murdered brother one last word also let a stranger clone your mother's voice to empty her savings.
The technology doesn't care which one it's used for. That part is up to us, and to whatever foundation we decide to build it on.
You are going to outlive yourself. The only real question is who gets to hold the copy.
