How AI avatars of the deceased damage their bereaved and real-world counterparts ethically, legally, and psychologically.
By Jasmine Chow, Associate Editor
Photo source: Ibraheem Lodhi
Over the past few years, technological development in the artificial intelligence (AI) sector has boomed. The emergence of large language models and generative AI capable of creating text, images, video, and sound has captured society. Recently, AI technology has progressed to be able to generate lifelike digital avatars of the deceased—termed “deathbots” or “griefbots”—using the digital data naturally collected over a person’s life. In this way, the deceased can almost seem to live on forever.
Avatars of the deceased aren’t exactly new, and neither are aspirations of immortality. Deathbots have existed for over a decade (though these older models often spat out gibberish or had animations that came across more “uncanny valley” than “technological marvel”). Celebrities like Tupac and Michael Jackson have had their likeness replicated by holograms on stage. Immortality has been a fixture of our cultural consciousness from at least the time of the ancient Mesopotamians and has since inspired the idea of digital immortality, like uploading your consciousness to the infinite possibilities of cyberspace. Our technology isn’t quite there yet, but in recent AI deathbots that can learn to mimic your appearance, voice, and speech using just a few videos of you, we’re certainly approaching something similar.
2wai, a company claiming to be “the world’s first social app for avatars and the human layer for AI,” went semi-viral in November 2025 for an advertisement depicting a woman and her son interacting with an AI deathbot of his late grandmother over the years. The intended idea behind these deathbots sounds noble: a digital resurrection to allow us infinitely more time to communicate with a loved one, to see their face and hear their voice. It sounds almost intriguing. But deathbots have been faced with controversy and criticism from ethical, legal, and psychological experts—and for good reason.
The Ethics of Dignity, Autonomy, and Unending Grief
Consider the following scenario proposed by philosopher Patrick Stokes: imagine that your hand is amputated but kept viable. It’s then hooked up to a hypothetical biological WiFi that connects to your brain, such that you can wiggle your fingers and point at the words of this article while the rest of your body is across the room. Your hand is clearly still part of you as a person, despite it no longer being part of your body.
Now what if your amputated hand was connected to both your brain and a computer? Your brain can still process sensations from the hand, but the computer is in control of your hand’s muscles, working off of an algorithm of how the computer thinks you would normally move your hand. You don’t directly control the computer, but it’s observed you and how you move and perhaps you’ve even fed it some information about your quirks (e.g. “I rub my thumb and forefinger together when I’m thinking”). Is your hand still a part of you then?
As Stokes predicted, you’re probably saying no. The key difference here is in the lack of autonomy and agency—in the case of deathbots, it’s the same thing. There is a difference between someone choosing to leave a piece of themselves behind after they die, like a video recorded for their family, and then that family taking that video and feeding it to a computer to turn their personhood into an AI deathbot. It’s a loss of dignity, autonomy, and personal agency for a person who has already lost these uniquely human things through death.
Moreover, deathbots are used not only as vessels for bereavement but also as digital commodities. Ethicists Carl Öhman and Luciano Floridi argue that “digital afterlife” services are incentivized to depict the deceased in their most consumable, profitable form rather than what is most accurate (though the 2wai app is currently entirely free to use, their website states that subscriptions and in-app purchases may be included in the future). This alters the deceased’s personal identity and further deprives them of the agency they once had. And is it ethical to profit—materially or socially—off of another person’s death or grief?
Now consider: what if you, of sound body and mind, decide to upload all your personal data to a deathbot so that your future great-great-grandchildren could forever hear your endless wisdom? You choose to do so while fully understanding the potential risks of your likeness being inaccurate or misused.
The issue here shifts to the ethics surrounding the bereaved. Researcher Nora Freya Lindemann writes that deathbots can affect their users in the intended ways, such as to help deal with grief, but can also make users reliant on deathbots (more on this later). The grief cycle is dysregulated; it is perpetuated rather than processed. But ethics aren’t the only issue surrounding digital resurrection.
Legality of Deathbots: Exploitation and Ownership
Because advances in AI technology have accelerated so quickly, lawmakers have struggled to catch up. Currently, as law professor Victoria J. Haneman puts it, the “digital right to be dead” has yet to be recognized as a legal right.
This issue exists at a complex intersection of technology law, privacy law, and property law. How can deathbot companies, and by extension, generative AI companies, be regulated and held accountable by our governments? How do we protect people’s data in a world powered by user information? And who owns the data of the deceased—Big Tech, the bereaved, the deceased, or someone else entirely?
Floridi argues that our personal identities exist “…as an informational structure; a narrative constituted by everything that defines it: memories, biometrical information, search history, social data, and so on. Thus, people do not merely own their information, but are constituted by it, and exist through it…We are our own information and our personal data are our informational bodies.”
In this view, it’s the deceased who own their data and have the right to keep it. But legally, it isn’t as simple. In the United States, the birthplace of many generative AI models, the dead have no right to privacy. Personal data is not considered to be property. Deepfake and AI avatar imagery are not criminalized, though there have been calls for legislation to protect the subjects of this imagery. Legally, there is little to nothing preventing the average person from generating a deathbot of their deceased family member, even if they had written “NO DEATHBOTS” in big red letters at the top of their will.
It’s clear that in this digital age, policymakers need to consider how to protect citizens against immoral AI usage and the loss of data privacy. Researchers have been proposing regulatory frameworks to protect both the living and the dead from AI deathbots—Öhman and Floridi propose one where digital remains are treated with the same respect as archaeological remains, and where digital afterlife companies must not commercially exploit the deceased—but none have yet been enacted into law.
Worse yet, a proposed Artificial Intelligence and Data Act died in Canadian Parliament after years of criticism and controversy. So unfortunately for now, the future of AI regulation in Canada is uncertain.
Psychology, Pseudo-bonds, and Playing Pretend
Another key aspect of deathbots to consider is the psychological impact that they can have on the bereaved. Deathbots, by design, require mutual interaction between the user and the deathbot; both can reply to the other, and the deathbot can create new messages on behalf of the deceased person they represent. This isn’t the case for pre-deathbot-era grieving, where the most one could do was to simply imagine what their deceased loved one would say.
This highlights a key difference between using deathbots and the traditional grieving process: a deathbot continues an external bond between the bereaved and the deceased, externalizing the privacy of grief, whereas the traditional grieving process involves changing a previously external bond to an internal one. Lindemann calls these deathbot relationships “pseudo-bonds,” highlighting the perpetuance of an emotional connection that no longer physically exists.
Lindemann goes on to describe that deathbots can create a unilateral emotional bond from the bereaved and impact the way they form relationships with other humans. Modern deathbots use generative AI to synthesize video and audio outputs that, to users, can almost appear real and human. Deathbots are also always available and answer mostly in expected ways. However, they lack true emotion and the ability to mutually bond with their users. The expectations and unilateral bonds formed by the bereaved through prolonged deathbot usage can bleed into interactions with real humans, preventing them from forming mutual relationships or causing surprise when a human doesn’t behave in the expected ways a deathbot would.
It’s important to note that pretending that a person is not really deceased and anthropomorphizing digital objects can happen naturally. But deathbots have the potential to unnecessarily extend the grieving period and create an emotional reliance from the bereaved on the deathbot; a person turning towards deathbots to ease their grief can prevent them from fully acknowledging death, leading them to continuously return to the deathbot for emotional comfort and safety.
A Personal Reflection on Deathbots
In 2017, Jang Ji-sung’s seven-year-old daughter, Na-yeon, died from blood-related diseases. Three years later, the Korean documentary Meeting You showed Na-yeon tearfully reuniting with her daughter using virtual reality technology and an animation constructed from photos and memories of the young girl. “At the end of the day, she said she was happy to see her daughter on a big screen, rather than feeling manipulated,” said Kim Jong-woo, the creator of the documentary.
Of course, this isn’t a one-to-one comparison with AI deathbots. Ji-sung could not interact directly with the animation with her daughter, nor did Na-yeon’s avatar have the capacity to answer any of her mother’s questions. The documentary was inevitably met with controversy and arguments over the exploitation of a child’s death. But for Ji-sung, who was happy to see her daughter one last time, the process was healing.
Technology can certainly have a positive impact on bereaved people. The difference between an animation of Na-yeon and an AI deathbot, however, is stark. Deathbots are uniquely interactive and accessible. They can be made for profit or commercial gain. Further, there are currently no regulations or long-term studies surrounding their usage or existence, allowing digital afterlife companies to run completely unchecked with unclear potential psychological effects. Perhaps one day in the future this will be changed. But for now, in this age of the AI boom, we have to be careful about deathbots and consider their dangers.
Over the time I spent researching and writing this article, I lost a family member. The month before that was the anniversary of the death of another. Not once did I consider uploading their likenesses to an insentient AI deathbot. It wouldn’t be fair to depersonalize the people that I’ve lost, to take away their agency and self-expression even more so than death already has. And it wouldn’t be fair to myself or to my family, who are left to pick up the pieces to patch the hole left behind after someone’s passing.
Grief is difficult. It’s heavy and comes in millions of forms, and it never truly leaves you; rather, it builds and wanes, builds and wanes, over and over again. But grief is simply a reflection of the love you held for someone: the memories you shared together and the presence they had in your life.
These memories are yours to keep and carry forward—not something that should anchor you in the past.
