Michael Bommer read 300 scripted phrases into a microphone, then let a two-day computing process turn the recording into an AI version of his own voice and personality, according to a June 2024 Fortune report. He was 61, a Berlin startup founder dying of colon cancer, and he was paying Eternos, a company run by his old colleague Robert LoCascio, 15,000 dollars, so that after he died his wife Anett and the rest of his family could still ask him things. Bommer got to say yes. Most people whose voices end up training a chatbot never get asked.
That is the quiet problem sitting underneath the fast-growing business some call the digital afterlife industry: StoryFile, HereAfter AI, Eternos, and a handful of newer entrants that train chatbots on a dead person's texts, voicemails, and social posts and hand the result back to the grieving as something you can still talk to, plus users of general AI-companion apps like Replika who repurpose them the same way. One widely cited market estimate, from The Business Research Company, puts this category at 35.8 billion dollars this year, growing to near 61 billion by 2030. NPR has separately reported estimates running as high as 80 billion within a decade; the spread says less about the industry's actual size than about how loosely "digital afterlife" is still defined. What is not in dispute is that the product exists now, people are buying it, and in most places the law has nothing to say about whether the person being cloned would have wanted that.
The public, it turns out, has an opinion even where the law does not. Masaki Iwasaki, then at Harvard Law School, ran a survey experiment asking American adults to judge a hypothetical: a young woman dies, a company offers to digitally resurrect her, and her friends have to decide. When the vignette said she had consented before dying, 58 percent of respondents called it acceptable. When it said she had explicitly refused, acceptability collapsed to 3 percent. People also turned out to be wary of their own resurrection: 59 percent said they would not want to be brought back even if they had agreed to it in advance. Consent, in other words, is close to the entire moral question for most people asked directly, and building a griefbot without it is something almost nobody thinks is fine.
Two researchers at Cambridge's Leverhulme Centre for the Future of Intelligence, Tomasz Hollanek and Katarzyna Nowaczyk-Basinska, tried to turn that instinct into something a company could actually build to. Their 2024 paper proposes what amounts to guardrails for the industry: get consent from both the person being simulated and the person who will talk to the bot, be honest at every turn that the user is talking to software, keep the product to adults, and build a dignified way to retire a deadbot rather than leaving it to run forever or vanish without warning. They floated a scenario they called MaNana, in which a grandchild builds a grandmother chatbot without asking the grandmother, and after the free trial ends, the bot starts working ads into the conversation. It read as a warning at the time.
Two years later, it reads like a memo the industry already has. Alex Quinn, the CEO of Authentic Interactions, StoryFile's parent company, told NPR in 2025 that he would not want a deadbot reciting scripted ad copy. "Why the hell would I want my grandma to say anything that's not authentic?" he said. "It would really turn me off as a consumer." But ruling out fake ad copy is not the same as ruling out advertising. Quinn said he remains "absolutely interested" in other ways to make deadbots ad-friendly, including working ads into a conversation the way a commercial break interrupts a TV show, or having the avatar quietly learn what a living user likes so an advertiser can use it. "We can instruct those avatars to actually probe for information," he said. "Who's your favorite athlete? What jerseys might be interesting to you?" Companies, he added, are going to try to make as much money out of AI avatars of both the dead and the living as possible. Right now there is no rule requiring anyone's consent to build the bot in the first place, let alone to decide later what the bot is allowed to ask you in order to sell you something.
Underneath the consent question sits a second, harder one: whether talking to a dead person's chatbot helps you grieve or keeps you from it. Nora Lindemann, whose 2022 paper "The Ethics of Deathbots" got here before most of the industry did, argues that grief is supposed to change precisely because you can no longer interact with the person who died; a chatbot that keeps answering interrupts that process rather than easing it. Ethicists including Jessica Heesen at the University of Tubingen have described these tools as acting like a painkiller, numbing a loss you still need to feel. Not everyone in the research agrees. A 2026 study of deadbot users presented at the CHI conference on human-computer interaction found some people experiencing comfort rather than harm: one participant described her deadbot in a published interview as "a crutch that can support me for a moment, making me feel as if Grandma is by my side."
None of this is settled, and that is the point. A product this intimate, this expensive in some cases and free in others, is being sold to grieving families faster than any legislature is writing rules for it. If you do not want to become someone else's chatbot after you die, or you do, say so in writing now, the same way you would with an organ donor card or a will. Right now, in most of the world, nobody is required to ask.
