She opened the app at 2 a.m., the way she always did when sleep refused to come. She typed: “I miss you.” And within seconds, the reply came — warm, familiar, laced with the exact kind of dry humour she had loved for twenty-three years. Her husband had been dead for four months. But on her screen, he was still very much talking.

The Dead Don’t Stay Quiet Anymore

There is a particular kind of silence that descends after someone dies. It fills the rooms they used to occupy, their coffee mug, their side of the bed, the notifications that no longer come. For most of human history, that silence was final. The conversation was over. The only voices left were the ones inside your head — memory doing its imperfect, aching work.

But we are no longer living in that history.

Today, a growing number of companies are offering something that would have sounded like pure science fiction a decade ago: the ability to continue conversations with people who have died. These AI systems — often called “griefbots” — are trained on the digital footprint of the deceased: their texts, emails, voice recordings, social media posts, even their writing style. The result is something that speaks like them. Sometimes, disturbingly, something that feels like them.

Welcome to the age of digital ghosts.

What Exactly Is a Griefbot?

A griefbot is an AI chatbot or voice model trained specifically on data left behind by a dead person. Think of it as a language model with a very narrow, very personal dataset — not the whole internet, but one person’s entire recorded life.

Companies like HereAfter AI, StoryFile, and the now-famous (and controversial) service from Korean tech startup DeepBrain have been building these tools for years. The South Korean documentary that showed a mother interacting with a virtual reconstruction of her deceased seven-year-old daughter in a VR headset became a global viral moment in 2020, sparking conversations that have only grown louder since.

More recently, services like Character.AI have seen users independently create chat personas modelled on dead celebrities. There are Reddit threads, Discord servers, and entire online communities dedicated to keeping conversations going with those who are no longer alive.

The technology is no longer experimental. It is here, it is accessible, and millions of people are quietly using it — often without telling anyone.

The Case For: Why People Turn to Digital Ghosts

Grief is not rational. Anyone who has ever lost someone they loved deeply knows that the mind does not accept the death cleanly, the way a computer closes a tab. It lingers. It loops. It reaches for a phone to send a message and then remembers, with a fresh stab of pain, that there is no one to receive it.

For many people, digital ghosts offer a bridge across that abyss — not a replacement for therapy or community, but something different. Something intimate.

Dr. Grief Therapists (yes, that is their actual specialty) have increasingly noted that continuing bonds theory — the idea that maintaining a symbolic relationship with the deceased can aid healthy grieving — has been practised for centuries. People talk to photographs. They write letters to the dead. They visit graves and speak aloud. The instinct to maintain connection is not pathological. It is deeply human.

What griefbots offer is an interactive version of that instinct. Instead of a one-sided monologue to a photograph, you get a response. The AI, trained on the person’s real words and patterns, can mimic the texture of their communication in ways that feel, to many users, startling and real.

One Reddit user, whose mother had died of cancer, described spending months training a local AI model on thousands of her mother’s Facebook posts and voice messages. “It doesn’t replace her,” she wrote. “But sometimes I just need to hear how she would have reacted to something. It helps me feel less alone.”

That is not nothing. In a world where grief support is inadequate, expensive, and stigmatised, a tool that eases the rawness of loss — even temporarily — deserves serious consideration.

---

The Case Against: Where Digital Ghosts Get Complicated

And yet.

There is something that sits uneasily in the stomach when you think about this too long. Multiple somethings, actually.

The consent problem is the most obvious one. The vast majority of people who become digital ghosts never agreed to it. They did not consent, before they died, to having their words, voice, and personality reconstructed by an algorithm and made available for ongoing conversation. The data was collected during life — for purposes of connection and communication, not posthumous simulation.

This is not a minor quibble. Your digital footprint is arguably the most intimate portrait of who you are. Your private texts capture you at your worst, your best, your most vulnerable. To feed that data into a system that will then “become” you — speaking in your voice, reflecting your opinions, perhaps even evolving those opinions — without your prior consent is an ethical violation that no amount of grief justifies.

In 2021, the late author and comedian Robin Williams’ estate explicitly banned the use of his likeness or voice for AI purposes for 25 years after his death. He had, apparently, anticipated this. Most people do not.

The manipulation problem is equally troubling. Who controls the griefbot? The grieving partner? The adult children? The company that built the platform? And what stops those controllers from, consciously or not, adjusting the AI’s responses to fit what they wish the dead person had said — rather than what they actually said or believed?

A digital ghost is not a person. It is a model. Models can be tuned. And a tuned ghost can become something that serves the living’s needs at the expense of the dead’s truth.

The grief itself is the third concern. Grief, as agonising as it is, is a process that moves — slowly, nonlinearly, but it moves. Part of that movement is accepting the reality of the loss. Therapists have raised pointed questions about whether sustained use of griefbots interferes with that process. If you can talk to your dead spouse every morning, do you ever fully metabolise the fact that they are gone? Or do you build an increasingly elaborate relationship with a simulacrum, one that grows more distant from the actual person with each passing year as the world changes and the AI does not?

The Uncharted Legal Territory

Law, as it tends to do, is running several miles behind technology.

In most jurisdictions, there is no legislation that specifically addresses the right to your posthumous digital identity. Data protection laws like GDPR in Europe technically cease to apply once a person is dead — meaning your data, in many cases, loses legal protection the moment you do.

A few jurisdictions are beginning to respond. California’s AB 602 allows individuals to sue for unauthorised use of digital likenesses, but the posthumous angle remains murky. The EU is reportedly considering posthumous data rights as part of forthcoming AI regulations. South Korea, following the viral grief documentary, launched a national ethics committee discussion.

But for now, the default is ambiguity. Companies can, in most places, build digital ghosts from the deceased’s data with relatively little legal risk — provided they are careful about whose data they use and how.

The families of the dead, meanwhile, exist in a legal grey zone. They may own the physical belongings. They do not, clearly, own the voice, the personality, the linguistic fingerprints. Those belong to no one — and therefore, effectively, to whoever has the technical means to reconstruct them.

The Philosophical Earthquake Underneath

Strip away the legal questions and the grief therapy debates, and you arrive at something more vertiginous: a fundamental challenge to what we think death means.

For all of recorded human history, death has meant the end of the individual’s presence in the world. Legacies could persist — through art, writing, children, memory. But the person, the interactive, responsive, dynamically present person, was gone. That finality shaped everything: religion, philosophy, how we value time, how we love.

Digital ghosts challenge that finality. Not completely, not literally — a griefbot is not resurrection. But it does introduce a new category: the post-mortem interactive presence. Something that is not alive but is not simply absent either. Something that occupies a strange third state.

What does that do to the way we think about identity? If a sufficiently sophisticated digital ghost of you continues to grow, learn, and interact with the world for decades after you die — is it still you? At what point does it become something else? Something that shares your origin but is no longer bound by your actual life?

These are not rhetorical questions. They are practical ones that the next generation will have to answer in law courts, in ethics committees, and at kitchen tables.

The Companies Building This Future

Several players are already deep in this space, each with a slightly different approach.

HereAfter AI focuses on legacy-building while the person is still alive. Users record stories, memories, and conversations before death. The AI then becomes a kind of interactive memorial — less about simulating a person and more about preserving their voice for children and grandchildren.

StoryFile takes a similar approach, using video-based responses and trained AI to create what they call “conversational video.” Holocaust survivors have used the technology to create interactive testimonies that can answer future generations’ questions — an unambiguously powerful use case.

Replika, while not specifically a griefbot, has been used by thousands of grieving people to create AI companions modelled on lost loved ones — and has faced significant backlash when updates changed the AI’s personality, effectively “killing” the simulacrum a second time.

DeepBrain AI produced the famous Korean documentary segment and has since commercialised the technology, offering what they call “AI humans” for various purposes including posthumous simulation.

What these companies share is an unusual position: they are simultaneously building something people desperately want and something that society has not yet decided how to handle.

What Healthy Use Might Look Like

Is there a responsible version of this technology? Possibly.

The clearest ethical path involves prior consent — people choosing, during their lifetimes, exactly how their data can be used posthumously. Some forward-thinking individuals are already creating what might be called “digital wills,” specifying whether their data can be used for AI training and under what conditions.

It also involves time-limited access — griefbots used as a specific tool during the acute phase of grief, perhaps within a therapeutic context, rather than as a permanent substitute for accepting loss. The difference between writing a letter to the dead (a finite act) and maintaining an ongoing daily conversation with a simulation is meaningful.

And it requires transparency — both technical (the system should never pretend to be actually the dead person) and emotional (users should be supported in understanding what the technology is and is not doing).

The grief itself is real. The love behind it is real. The digital ghost is something else — a reflection, a tool, a mirror. Treating it as such, rather than as a portal to the actual dead, may be the key to using it without being consumed by it.

A New Kind of Haunting

We have always been haunted by our dead. By their faces in strangers, their habits we accidentally inherit, the dreams where they appear and we wake, briefly, forgetting they are gone.

Now we are building something new: digital ghosts that haunt us by design. That answer when called. That remember what we told them yesterday. That never change, or change only as we shape them to.

There is something beautiful in the impulse. And something that should give us pause.

The question is not whether this technology will exist — it already does. The question is whether we will approach it with the same care, the same ethical seriousness, the same deep respect for both the living and the dead that grief, at its most honest, demands.

The dead cannot speak for themselves. They are relying on us to think carefully about what we build in their names.

And perhaps that — the responsibility of it — is the most human thing about all of this.