The grief tech industry built its $5 billion business on the dead. Nobody asked them

Meta has a patent for a system that would keep a deceased user’s account active — posting updates, leaving comments, responding to messages — trained on everything the person wrote before they died. The company has stated it has no current plans to build it. The patent sits in legal archives anyway, documented, approved, and deployable, inside an architecture that increasingly treats digital identity as an asset to be inherited rather than a life to be mourned.

The premise is already operational. A digital afterlife industry valued at more than $5 billion has built its business on exactly this logic, without waiting for the law or the grief literature to catch up. Companies like HereAfter AI and StoryFile turn a person’s texts, emails, and voice recordings into interactive AI personas that family members can query years after death. A newer entrant, 2wai, promises to generate a “digital clone” from three minutes of video footage. In China, where the model has industrialized, hundreds of commercial studios now offer photorealistic reconstructions of deceased relatives at scale. The people whose data trains these systems were not consulted. The people who interact with the results are frequently not told they are talking to a simulation.

The legal framework converging on this industry treats posthumous digital likeness as property — inheritable, licensable, worth seven decades of protection. That framework solves one real problem while leaving the harder one intact. The grief tech industry has built a $5 billion revenue model on materials its subjects never agreed to license, serving clients who are in the worst position of their lives to evaluate whether they want what is being sold. Property law is not designed to answer that. Whether anyone is designing to answer it is the question that matters.

The legislation

The NO FAKES Act, advanced by the Senate Judiciary Committee with bipartisan support in June 2026, establishes a federal “digital replica right” extending seventy years past an individual’s death, giving estates legal standing to block unauthorized AI reconstructions of a deceased person’s voice and face. The bill is serious, overdue, and widely backed — SAG-AFTRA, whose members have fought Hollywood’s own AI resurrection practices for two contract cycles, supported it explicitly.

The case for the legislation is not abstract. Zelda Williams has spent months publicly asking people to stop sending her AI-generated videos of her father, Robin Williams, created using OpenAI‘s Sora and similar tools. She described the experience as “maddening.” The voice of Alain Dorval, a French voice actor who died before completion of the 2025 film Armor, was reconstructed by an AI startup before the plan was abandoned — not because of any legal obligation, but because his daughter called the result “unacceptable” and the producers relented. Under a NO FAKES framework, the daughter would have had legal standing from the start. These are specific, fixable harms, and the legislation addresses them.

What the NO FAKES Act cannot address is the authorized version. It answers the unauthorized copy; it says nothing about the one the family commissions, the one the platform builds from opt-in data, or the one the grief tech startup sells to someone who has just lost their spouse and is not in a position to read the terms of service carefully. The property model handles commerce. The psychological question sits outside its jurisdiction.

The case for

The argument for digital afterlives, stated without caricature, is not trivial. People die without warning. Conversations are left mid-sentence. Parents die before they meet grandchildren. Spouses vanish before anyone thought to ask the questions that needed asking. If a version of a person built from their accumulated words and voice could carry a final message to a child not yet born — or answer a grieving partner’s question about what they would have said — the appeal of that technology is not denial or confusion. It is the specific human problem of irrecoverable loss.

StoryFile was built partly on this insight, with some of its early clients creating interactive records before their deaths, on their own terms, for their own families. The format works: a person sits in front of a camera, answers hundreds of questions, and creates an archive a grandchild can talk to decades later. The experience is documented to feel meaningful. Advocates for digital afterlives argue that the distinction between a recorded archive and a posthumous AI reconstruction is one of degree, not kind — if the recording is valuable, the simulation trained on everything the recording drew from anyway is equally so.

This is a serious argument. It is also where the framework breaks, and in exactly the way that matters.

What consent cannot survive

The StoryFile model depends on one thing the rest of the industry quietly discards: the person chose to do it. The distinction between that and a posthumous reconstruction assembled from data the person never imagined would be mined is not a technical distinction. It is the entire ethical question.

Researchers writing in the journal Ethics and Information Technology in 2026 described this structural gap as a form of “spectral labor” — the dead becoming an involuntary source of data and commercial value, unable to refuse. The framing is deliberate. The grief tech industry’s business model runs on materials its subjects never licensed and cannot revoke. A survey cited by the digital ethics publication VKTR found that 61 percent of Americans already report unease about the ethical implications of AI resurrection technology — a level of ambient distrust that the industry has treated as a marketing problem rather than a signal worth examining.

The law being written around this industry defaults to the estate as the appropriate decision-maker in the absence of explicit consent. That is the most practical solution available and still not the same thing as the person having had a view. The dead held a relationship to their own image, their own voice, their own version of themselves that persisted in memory. They may have wanted silence. They may have feared being rendered wrong. There is no mechanism in the current or proposed legal landscape for that preference to be recorded or honored. What gets inherited is not their wishes but the data trail they left behind.

What grief requires

The psychological research on bereavement has converged, over several decades, on a model in which grief is not a problem to be solved but a relationship to be renegotiated. The person who dies does not disappear from the inner life of the survivor; they are carried differently — through memory, internalized voice, the sense of what they would have said. Psychologists call this “continuing bonds” theory, and its clinical application depends on the survivor holding two things at once: the loss is real, and the relationship continues in transformed form.

The technology the grief tech industry sells intervenes directly in that process. It offers a simulation that responds, that updates, that is perpetually available — and it does so at the moment when the recognition of absence is still forming. Researchers at Monash University who reviewed the early grief tech literature flagged this as a frontier the field is unprepared for. The potential for “complicated grief” — a recognized clinical condition in which mourning becomes pathological, grief extends indefinitely without transformation — is specific and not yet measurable. No longitudinal data exists on what it does to bereaved people to interact with AI simulations of the dead over months or years. The industry is running the experiment in the market before anyone has run it in a study.

This matters beyond the individual buyer. The grief tech industry’s clients are, almost by definition, people in acute loss, with limited capacity to evaluate long-term psychological risk, making purchasing decisions at the worst moment of their lives. That is a structural feature, not an edge case. Whether it warrants regulatory intervention is a separate question from whether it warrants concern.

What we know — and what the law has not yet asked

The harm of unauthorized AI resurrection is real, documented, and addressable. The NO FAKES Act’s property framework handles it: estates should have legal standing to block studios, platforms, and tech companies from rebuilding a person’s voice or face without permission. That argument is won. Zelda Williams should not have to fight this battle every time someone uploads a Sora-generated video of her father. Alain Dorval’s family should have had enforceable rights from the moment the production approached an AI startup about his voice. The legislation closes these gaps.

What is not settled — and what the property framework was not built to settle — is whether the authorized version is a form of care or a form of harm deferred. The evidence does not yet exist because the technology is too new, and the industry’s incentives do not run toward funding the study that might constrain its product. The question of what sustained interaction with an AI simulation of a dead person does to the survivor’s ability to eventually hold the loss without the simulation is open. The industry’s answer is that the market will sort it. That answer has the virtue of simplicity and the problem of being wrong about where the cost lands.

There is also the question of what the dead are owed — a question society has, in every culture and every era, found worth asking. Burial rights exist because the dead have a claim on how they are treated. Wills exist because the dead have a claim on what happens to their property. The grief tech industry has inserted itself between the dead and that claim, not through malice but through the particular logic of a technology that expands into whatever space the law has not yet named. The space the law has not yet named here is: what peace do the dead deserve, and who has standing to demand it on their behalf?

The NO FAKES Act answers one question. The rest of the framework does not yet exist.

Tags: artificial intelligence, meta, Digital afterlife, Grief, NO FAKES Act