Digital Immortality: Between Vision and Ethical Responsibility

In November 2025 an American company launched an app that generates a video persona of a person who has died from a few minutes of footage. Its launch video, a mother talking with the son she had lost, drew tens of millions of views, and within a week became shorthand for what must not be done in this field. We build a product in the same category. So the question "is this ethical" is not theoretical for us, it is the question we are asked in every conversation, and rightly so. In this article I answer it through three smaller questions, and three cases that actually happened.

Question 1: Consent, this person never asked for it

It is the strongest argument against the whole category, and there is no point dodging it: the person who died gave no explicit consent to a persona being built in their style. In 2024 two researchers at Cambridge's Leverhulme Centre for the Future of Intelligence published a call for safeguards, and coined a term that stuck: "unwanted hauntings" an AI persona of the dead that keeps reaching out to people who did not ask for it, or is used for purposes the person would never have approved (Cambridge's announcement).

What we do with that: the consent we cannot get from the person, we require from the family. A profile is created only after its creator confirms that close family members share the decision; the persona is built only from materials the family itself uploads, not from scraping the web, not from public posts; and it never reaches out to anyone on its own. And what we cannot do: verify that the family consent is real. We rely on whoever creates the profile, and we say so out loud.

Question 2: Rights, whose digital identity is it, and what happens when the company disappears

In July 2026 HereAfter AI shut down, the best-known product in the category, the one every article about "talking to the dead with AI" quoted. Users who had recorded hours of interviews with a parent found a website returning errors and an email address for enquiries. Some had exported their recordings in time; some had not (we wrote about it at length). The question "whose memories are these" suddenly got a very practical answer: whoever can download them.

And in Israel, where we are based? As of 2026 there is no dedicated law on the "digital identity" of a person who has died. What applies rests on the Privacy Protection Law and on inheritance law, both written before this question existed, both silent on who decides whether to build a persona of Grandpa. Which means the responsibility sits with the platform. Our position is simple, and written on our ethics page: the profile belongs to the family that created it; everything can be exported at any moment, in open formats; and everything can be deleted, including the reconstructed voice at the external voice provider, with no "cooling-off period" and no copy left behind.

Question 3: Harm, does it help grieving, or prevent it?

Honesty requires saying: the research is young, and there is no definitive answer. Early studies from 2026 found that for some mourners, a conversation with a persona of their loved one brought comfort and a kind of closure; the same studies also document dependence, and the pain of an answer that is "not like them". We gathered everything that is known, with links, in Digital Legacy and Grief.

What can be done in the meantime is to build the boundaries into the product rather than leave them to policy. With us: the persona always identifies as AI: it does not "feel like them", it does not encourage you to forget; it does not discuss the death or its circumstances, only the life; there is no memory between conversations: every conversation starts fresh, on purpose, so that no accumulating "relationship" grows that cultivates dependence; and it does not invent: if something is not in the materials, the answer is "I don't remember", not a memory that never was. These boundaries are fixed. The family cannot switch them off either, and that is deliberate.

What we still do not know

• There is no halachic ruling on this, and we do not pretend to offer one. A family that wants guidance, better before creating the profile.
• There is no Israeli regulation, and no sign it is coming. Until it does, the rules above are what there is.
• There is not enough long-term research. What happens to a family after five years with a persona like this, nobody has measured yet.

And why at all

I built this product after losing my sister on October 7th. I went through every app in the category as a family member, not as a competitor, and what frightened me was not the technology, it was the promises. "Talk to her again" is a promise no product can keep, and whoever makes it is selling something else. I prefer a product that sometimes says "no" not about the death, no invented memory, no pretending, over a product that promises everything. If that means the persona sometimes answers "I don't remember" that is fine. That is how Grandpa would have answered too.

FAQ

What an AI avatar actually is, and what it is not, in Talking to a Deceased Loved One Through AI. The full boundaries, and how to measure us against them, on our ethics page.