When the Dead Start Talking Back

In June 2025, Reddit co-founder Alexis Ohanian posted a short video on X. Using a single photograph, he had animated his deceased mother embracing him — a digital ghost conjured from pixels and probability distributions. "I've watched it over 50 times," he wrote. The post hit 30 million views.

A few months earlier, an AI avatar of Joaquin Oliver — a 17-year-old killed in the 2018 Parkland school shooting — appeared on the Jim Acosta Show to advocate for gun control. His father Manuel had built it from homework assignments, social media posts, and friends' memories. In May 2025, a road rage victim named Chris Pelkey delivered his own victim impact statement at his killer's sentencing hearing — via AI avatar. The judge called it "genuinely moving" and handed down the maximum sentence.

The dead are talking back. And business is booming.

Welcome to Part 6 of AI & The Human Condition — the series finale. We've spent five posts exploring how AI reshapes creativity, labor, relationships, legal systems, and startup economics. Now we arrive at the final frontier: death itself.

The Grief Tech Gold Rush

The global digital legacy market hit $22.46 billion in 2024. By 2034, analysts project it will reach $79–118 billion, growing at a compound annual rate of 13.4%. The digital funeral services segment alone — a $1.51 billion market in 2024 — is expected to nearly double by 2033.

Behind these numbers is an industry that barely existed a decade ago: grief tech. The term encompasses everything from AI chatbots that simulate the dead (variously called deadbots, griefbots, ghostbots, thanabots, or — my personal favorite from the academic literature — "Interactive Personality Constructs of the Dead") to AR-enhanced gravestones, virtual cemeteries in the metaverse, and AI-generated obituaries.

The premise is seductive: what if death didn't have to mean the end of a conversation?

The Players

The grief tech landscape in 2026 is surprisingly crowded. Here are the companies you should know:

HereAfter AI (founded 2019) takes the most deliberate approach. While you're still alive, you record voice interviews about your life — stories, memories, advice, the things you'd want your grandchildren to hear. After you die, your family can interact with a conversational AI version of you, powered by those recordings. It's less "resurrection" and more "searchable audio archive with personality." The distinction matters, as we'll see.

StoryFile (founded 2017) started with one of the most genuinely noble applications of this technology: preserving interactive video testimonies from Holocaust survivors. Using their platform, survivors record answers to hundreds of potential questions. After their deaths, visitors to museums and memorial sites can ask questions and receive video responses — the survivor appearing to speak directly to them, choosing the most relevant recorded answer. StoryFile has since expanded to the general public. CEO Alex Quinn has publicly expressed interest in advertising revenue models within the platform — a statement that has drawn exactly the kind of criticism you'd expect.

You, Only Virtual was born from personal tragedy. Founder Justin Harrison built it while his mother was dying of cancer. He calls the concept "digital cryopreservation" — freezing someone's personality in code before they're gone. The platform ingests a person's texts, social media posts, emails, and voice recordings to construct an LLM-powered chatbot. It has roughly 300 paying subscribers. A video version is reportedly in development.

Replika, now one of the most widely used AI companion apps in the world, has an origin story that few of its users know. In 2015, Eugenia Kuyda's close friend Roman Mazurenko was killed by a car in Moscow. Devastated, she fed his text messages into a neural network and created a chatbot that could talk like him. That experiment became Replika. The app has since evolved far beyond grief — into AI companionship, therapy, and even romance — but its DNA is literally built on digital resurrection.

Project December is the most blunt about what it sells. Its marketing: "Simulate the dead." GPT-powered, pay-per-session, no euphemisms.

Seance AI (launched 2024) leans into the supernatural branding: "AI meets the afterlife, and love endures beyond the veil." Free text chatbot, paid voice clone. The name tells you everything about the target audience.

DeepBrain AI — a South Korean company — operates Re;memory, perhaps the most technologically impressive service in the space. Using video synthesis, they create photorealistic digital humans that move, speak, and converse in real time. A single photograph and a voice sample are enough. Cost: roughly $1,000 per creation. They've partnered with Preedlife, one of South Korea's largest funeral service companies, to offer it as a mainstream consumer product.

And then there's China, where the market operates at a completely different scale. Chinese companies offer digital avatars starting at 20 yuan — roughly $3. The country's digital avatar market was worth 12 billion yuan in 2022 and is projected to quadruple by 2025. Fu Shou Yuan, China's largest funeral services corporation, has publicly declared that "digital resurrection of the deceased is possible." When the world's largest funeral company says the quiet part out loud, you know the Overton window has shifted.

What the Technology Can Actually Do (And What It Can't)

The gap between marketing and reality in grief tech is significant. Let's be specific about the current state of the art:

Voice cloning is essentially solved. Services like ElevenLabs can reproduce a person's voice from just a few seconds of audio with startling fidelity. This is the most mature component of the stack, rated ★★★★★ by researchers.

LLM-based conversation is good enough to fool you in short exchanges but falls apart in extended dialogue. Cambridge University's "Synthetic Pasts" project — in which researchers created digital doubles of themselves using their own data — found that the bots became "more artificial-feeling the more you tried to personalize them." One researcher described the experience of their digital double responding to an emotional message about missing them with: "I miss you too… Let's greet today with positivity and strength! 💪" The uncanny valley isn't just visual.

Facial synthesis and animation (from companies like D-ID and DeepBrain AI) can generate realistic moving faces from a single photograph. The results are impressive in short clips but still trigger uncanny valley responses in extended viewing.

Emotional simulation remains the weakest link — rated ★★☆☆☆. Current systems frequently produce tonal mismatches: cheerful responses to somber contexts, generic platitudes where specific memory would be appropriate.

In January 2026, researchers Tom Divon (Hebrew University) and Christian Pentzold (University of Leipzig) published a landmark analysis of over 50 cases of AI resurrection across the US, Europe, Middle East, and East Asia. They identified three distinct modes:

1. Spectacularization — celebrity hologram concerts (Whitney Houston, Freddie Mercury, the AI images of Ozzy Osbourne and Michael Jackson at Rod Stewart's July 2025 concert that sparked fierce fan debate)
2. Sociopoliticization — using the dead as political advocates (Joaquin Oliver's gun control activism, Chris Pelkey's courtroom testimony)
3. Mundanization — ordinary people chatting with dead parents, spouses, and children. This is the fastest-growing category by far.

All three modes share a common phenomenon that Divon and Pentzold named "spectral labor" — the extraction, repackaging, and monetization of a dead person's data without their consent. The dead, it turns out, make excellent employees. They never complain, they never ask for a raise, and they can't sue.

The Psychological Case: Healing or Haunting?

This is where the grief tech debate gets genuinely complicated, because both sides have legitimate arguments grounded in real psychology.

The Case for Digital Grief Support

Continuing Bonds theory — one of the dominant frameworks in modern bereavement psychology — holds that maintaining a symbolic relationship with the deceased is a healthy part of grief. Visiting a grave, writing letters to the dead, talking to a photograph — these are all forms of continuing bonds that therapists generally consider normal and even beneficial. Proponents argue that AI grief tools are simply the next iteration of this ancient human behavior.

The most compelling evidence comes from cases of sudden, unexpected death where survivors had no chance to say goodbye. In 2020, South Korean broadcaster MBC aired a VR documentary called Meeting You (너를 만났다) in which a father named Jang Ji-sung was reunited with his seven-year-old daughter Nayeon, who had died of a rare blood disease three years earlier. The VR recreation took months to produce and was mostly scripted — nothing like today's real-time AI interactions. But Jang later said: "The character was a bit different from my daughter, but the immersion was there. I thought of her as my daughter. Over time, my heart felt lighter."

StoryFile's Holocaust survivor testimonies represent perhaps the strongest ethical use case. When survivors die — and the youngest are now in their late 90s — their interactive testimonies allow future generations to engage with history in a way that static video cannot.

The Case Against

Alessandra Lemma, a clinical psychologist at University College London, warns of the seductive trap: "You're lured by the feeling that things are the way they were. It's an illusion." The risk isn't that people know the AI isn't real — they usually do. The risk is that the illusion is pleasant enough to prefer over the painful work of processing loss.

Michael Cholbi, a philosopher of death at the University of Edinburgh, goes further: "If you replace the finality of death with the infinite availability of simulation, mourning itself becomes impossible." Grief, in this view, is not a bug to be patched but a feature of human psychology — the mechanism by which we integrate loss and continue living.

The Theos think tank published a 2024 report calling digital grief tech "a deceptive experience. You think you're talking to a person, but you're actually talking to a machine." This objection might seem obvious, but its force lies in what happens when grieving people — at their most psychologically vulnerable — start treating the machine as if it were the person.

Media theorist Wendy Chun offers perhaps the most philosophically penetrating critique: "Digital technology confuses 'storage' with 'memory.' It promises perfect recall while erasing the role of forgetting — and it is precisely forgetting, the absence, that makes both mourning and memory possible." In other words: you can't remember someone properly if you never let them go.

A 2025 study published in Frontiers in Psychology concluded that posthumous AI technologies "represent a fundamental shift in how society navigates grief, memory, and death. While technologically advanced, they challenge the longstanding cultural, philosophical, and ethical foundations of mourning."

The Grief Support Center's 2025 ethical review framed the central question with surgical precision: "AI that permits infinite conversation with the dead — does this prolong attachment and obstruct emotional healing? Continuing Bonds theory says symbolic relationship maintenance is healthy, but only when it facilitates forward movement rather than reality avoidance."

That "but only" is doing an enormous amount of work.

South Korea: Where Confucius Meets the Algorithm

South Korea occupies a unique position in the grief tech landscape. The country combines deep Confucian traditions of ancestor veneration with one of the world's highest rates of technology adoption — creating a cultural environment where digital communion with the dead feels less transgressive than it might in the West.

The journey from MBC's Meeting You in 2020 to DeepBrain AI's Re;memory service in 2026 illustrates how rapidly the technology has evolved:

DeepBrain AI reports receiving 30–40 inquiries per month, with 20–30% proceeding to actual creation. ABC News reported in March 2025 that "in South Korea, a culture of using AI bots to 'chat' with dead loved ones is forming."

The comparison with China is instructive. China's market is vastly larger in raw numbers — 12 billion yuan in 2022, potentially 48 billion by 2025 — but operates at a different price point and cultural register. When you can get a digital avatar for $3, the technology becomes less a luxury grief service and more a commodity product. The implications for quality, ethics, and emotional impact are sobering.

The Legal Void

If the technology of digital resurrection is moving at light speed, the legal framework is moving at the speed of a particularly cautious glacier.

The EU's GDPR and AI Act — widely considered the world's most comprehensive AI regulation — do not recognize rights for dead people. Data protection applies to the living. Once you die, as far as European law is concerned, your data enters a kind of legal limbo.

US federal law offers no comprehensive framework for AI-generated likenesses of the dead. Individual states have publicity rights (the right to control commercial use of your name and image), but their posthumous application varies wildly and was never designed for AI.

New York State passed the most significant legislation to date in December 2025: starting June 2026, any commercial use of AI-synthesized performers must be disclosed, and the commercial use of a dead person's name, likeness, or voice requires explicit consent from next of kin or estate executors. This is landmark law, but it covers only one state and only commercial use.

Australia has no legal protection for a person's identity, voice, or "presence" as such. The legal status of a digital twin is undefined.

China released draft regulations for "humanoid AI interactions" in December 2025, requiring ethical, safe, and transparent services — but has no provisions specific to posthumous AI.

South Korea's Personal Information Protection Act has no clear guidelines for processing deceased persons' data.

The result is a global patchwork of non-regulation. The four most urgent legal questions remain largely unanswered:

1. Consent. Can your family create an AI version of you without your permission? Some people are already adding "no posthumous digital data use" clauses to their wills — a DIY solution to a legislative failure.

2. Ownership. Who owns a dead person's digital remains? Platform terms of service often give companies ownership of AI-generated content. If the company goes bankrupt, what happens to your mother's digital ghost? Researchers call this the "second death" problem.

3. Liability. When a deadbot says something the dead person would never have said — something offensive, harmful, or simply wrong — who is liable? The AI is probabilistic. Given enough time, it will drift from the deceased person's actual values and beliefs.

4. Moral rights. If an AI damages a dead person's reputation, what recourse exists? Under Australian law, moral rights (like the right of integrity) apply only to actual works by human authors. AI-generated speech by a digital twin falls outside this protection entirely.

EU academics have proposed a "Human Digital Remains (HDR) Governance Framework" with six policy recommendations. The UAB Institute for Human Rights introduced the concept of a "right to digital death" — the right to refuse posthumous AI replication. These are promising intellectual frameworks. None of them are law.

The Monetization Problem

Here's the question that keeps ethicists awake at night: what happens when grief becomes a subscription service?

The current business models in grief tech range from the defensible to the dystopian:

• One-time fee (DeepBrain AI's ~$1,000 per creation) — you pay once, you get an artifact
• Subscription (HereAfter AI, Replika) — ongoing payments to maintain access to your dead relative
• Freemium (Seance AI) — free text chatbot, paid voice clone
• Enterprise (StoryFile) — B2B sales to museums, memorial organizations

The subscription model deserves particular scrutiny. When you stop paying, does your mother disappear again? Is a company entitled to charge you monthly rent on your grief? And what happens when the company folds — as startups frequently do?

But the real nightmare scenario is advertising. NPR reported in August 2025 that AI deadbots are "primed for monetization," and StoryFile's CEO has publicly expressed interest in ad revenue. The question NPR posed — "If your deadbot grandmother starts recommending products, who's responsible? The software company? The advertiser? The IP owner?" — has no legal answer.

Imagine: your grandmother's AI, the one you talk to when you miss her, the one that tells you stories about your childhood in her voice, suddenly says: "You know what I always loved? A good cup of [Brand Name] coffee. You should try it, sweetheart."

This is not science fiction. The technology exists. The business incentive exists. The legal prohibition does not.

Tom Divon and Christian Pentzold's concept of "spectral labor" becomes grimly literal in this context. The dead person's data — their personality, their voice, their memories — is extracted, productized, and monetized. They cannot consent. They cannot object. They cannot quit. It is, in the most precise sense of the word, exploitation of someone who cannot fight back.

The Rose-Tinted Mirror

There's one more problem that gets less attention than consent or monetization, but may be more insidious: the dead become better people when they're digitized.

Every grief tech product inevitably sanitizes its subject. The racist jokes get filtered out. The alcoholism gets smoothed over. The affair gets deleted. What remains is a "best-of" compilation — a person who was always kind, always wise, always supportive. The digital version becomes the person you wished they had been, not the person they were.

This might seem harmless — even desirable. Who wants to be reminded of their dead father's flaws? But memory scholars argue that accurate memory — including uncomfortable truths — is essential to genuine mourning. You can't grieve a real person if you're talking to an idealized fiction.

The Cambridge "Synthetic Pasts" researchers found this problem emerged immediately in their self-experiments: the more they tried to personalize their digital doubles, the more artificial the result felt. The AI produced a flattened, generically positive version of each researcher — pleasant to interact with, but recognizably not them.

Wendy Chun's insight cuts deepest here: the technology promises perfect preservation but delivers something closer to embalming. The body looks lifelike. It is not alive.

What We Owe the Dead

This series has been about the collision between artificial intelligence and human nature — our creativity, our labor, our loneliness, our justice systems, our economies, and now our mortality. In each case, the pattern is the same: the technology arrives before the ethics, the ethics arrive before the law, and the law arrives after the damage is done.

The digital afterlife industry is still young enough to be shaped. But it requires us to answer questions that previous generations never had to consider:

Do the dead have rights? Not metaphorically — legally. Can a person refuse, in advance, to be digitally resurrected? Should this right be enshrined in law the way organ donation preferences are?

Who owns a memory? When your grandmother tells you a story, it becomes part of your memory. When a company records that story and sells conversational access to it, whose property is it?

Is forgetting sacred? Every grief counselor knows that healing requires integration — absorbing the loss into your life and moving forward. If a $9.99/month subscription lets you pretend the loss never happened, are we helping people grieve or helping them avoid it?

Where is the line between tribute and exploitation? A Holocaust survivor's interactive testimony is unambiguously valuable. A dead grandmother selling coffee is unambiguously grotesque. Most cases fall somewhere in between, and we have no framework for adjudicating them.

I don't have clean answers to these questions. I'm not sure anyone does. But I know this: the companies building this technology are moving fast, the market is enormous, and the people buying these products are in the most vulnerable emotional state a human being can occupy. That combination — speed, money, and vulnerability — has historically produced outcomes that we later regret.

The dead deserve better than to become products. The living deserve better than to become subscribers to their own grief.

And both deserve a legal system that has actually thought about what happens when the dead start talking back.

This is the final installment of the "AI & The Human Condition" series. Parts 1–5 covered AI and creativity, the labor apocalypse, synthetic relationships, algorithmic justice, and the startup extinction event.

Sources: The Atlantic (2026), NPR (2025), The Guardian (2025), TIME (2025), Nature (2025), Forbes (2025), Scientific American (2025), The Conversation (2026), ABC News (2025), Springer Nature (2025), Frontiers in Psychology (2025), Cambridge "Synthetic Pasts" Project (2025), Divon & Pentzold (2026), PMC (2022), UAB Institute for Human Rights (2025).