What happens when a digital legacy company shuts down, and what you are owed
When a digital legacy company shuts down, your recordings do not transfer anywhere. They survive only if you exported them first, in a standard format, before the notice appeared. HereAfter is closing now, and StoryFile filed for bankruptcy in 2024. The memories were never the fragile part. The company holding them was, and that is the part almost nobody evaluates before they upload a parent's voice.
For a few years the whole debate about preserving a person with AI has fixated on one question: can a machine really hold who someone was? It is the interesting question, the one that fills magazine features and dinner tables. It is also the wrong one to lead with, because it quietly assumes the machine will still be running. The recordings are only as permanent as the smallest, least funded thing in the chain, and that thing is usually not the model or the servers. It is the company. Companies are mortal in a way that memories are not supposed to be.
What actually happens to your recordings when a legacy company shuts down?
Whatever the shutdown notice allows, is what happens. There is no escrow, no regulator, no automatic handoff to a successor who has promised to keep your father talking. When a company in this category closes, your data does not move to safety on its own. It sits on infrastructure that someone has stopped paying for, behind a login that will eventually stop resolving, and the only bridge out is whatever export the company built while it was still alive and still cared to.
Look at how it is playing out right now. HereAfter AI, for years the most recognizable name in the record-your-memories category, is closing. As of this writing its homepage carries a single farewell line: "It's been an honor to help families record their memories for loved ones, but unfortunately HereAfter is shutting down" (hereafter.ai, retrieved 19 July 2026). Account holders are told to email support@hereafter.ai to retrieve their recordings, with no deadline posted. The company's /faq and /blog pages already return errors, and its app domain, hereafter.app, now redirects to a "for sale" parking page. Somewhere in that quiet wind-down are families who recorded a parent who has since died, and whose entire path back to that voice is now a support email and a hope that someone still reads the inbox.
This is not one company's misfortune. StoryFile, which built interactive video "conversations" with recorded people and was demonstrated at high-profile memorials, filed for voluntary Chapter 11 bankruptcy on 5 May 2024 (case 7:24-bk-22398, U.S. Bankruptcy Court, Southern District of New York; PacerMonitor). Its assets were acquired by Key 7 Investment Company, and it announced emergence from bankruptcy on 3 March 2025 (AV Interactive, 3 March 2025). Read that arc carefully, because it is the good outcome, and it is still unsettling: your loved one's archive changed owners in a bankruptcy court, and the people who now hold it are not the people you trusted when you uploaded it.
Then there is Eternos, which marketed itself around digital immortality and was covered as such. In November 2025 it rebranded to Uare.ai and raised a $10.3M seed round from Mayfield and Boldstart Ventures, repositioning from posthumous legacy toward a personal AI for living professionals. Its founder, Robert LoCascio, told TechCrunch (11 November 2025) that most prospective users "weren't preparing for death." The company did not vanish. It simply decided its future was with the living, which is a rational business call and a quiet lesson for anyone who bought the immortality pitch: the company's mission is a variable, and yours is not the only vote.
The recordings did not fail. In every case the company changed, and the family found out afterward.
None of this is exotic to technology. The broader web has rehearsed it for a decade. Google Reader shut on 1 July 2013 (TechCrunch, 13 March 2013), with a narrow window to export before it went dark. Consumer Google+ deleted all its content on 2 April 2019 (TechCrunch, 30 January 2019), a mass erasure run by one of the most durable companies on earth. Vine closed in January 2017 (CNN, 17 January 2017). Every one of those gave users an export window, and every one of those windows closed. Products die on a schedule the user does not set. The only difference with a legacy company is the cargo. A dead feed reader took your unread articles. A dead legacy company takes the last recording of someone's mother laughing.
The digital legacy graveyard, by the numbers (as of July 2026)
A snapshot of the category's mortality, each item sourced and dated so you can check it yourself:
HereAfter AI is shutting down. Its homepage carries a farewell notice, recordings are retrievable only by emailing support, no end date is posted, its /faq and /blog pages return errors, and its app domain redirects to a "for sale" parking page (hereafter.ai, retrieved 19 July 2026).
StoryFile filed Chapter 11 bankruptcy on 5 May 2024, had its assets acquired by Key 7 Investment Company, and announced emergence on 3 March 2025 (PacerMonitor; AV Interactive, 3 March 2025).
Eternos rebranded to Uare.ai and raised $10.3M in seed funding, pivoting from posthumous legacy toward a personal AI for living users (TechCrunch, 11 November 2025).
Wider precedent for products orphaning user data: Google Reader closed 1 July 2013, consumer Google+ deleted all content on 2 April 2019, and Vine shut in January 2017. Each offered an export window that then expired (TechCrunch, 2013; TechCrunch, 2019; CNN, 2017).
Three companies, three different endings, closure, bankruptcy-and-acquisition, and a pivot to the living, and a single shared fact underneath. In none of them did the customer get a vote on what happened to the archive. The academic and press coverage has caught up to the stakes: Nature ran a feature titled "Ready or not, the digital afterlife is here" (Nature, 15 September 2025), and Scientific American surveyed the benefits and risks of chatting with recreations of the departed (Scientific American, 18 November 2025). The phenomenon is now mainstream. Its failure mode is still an afterthought.
Isn't this just how the technology industry works?
Yes, and that is the honest counterargument, so let me make it at full strength before I answer it. Most startups fail. Preserving data forever is genuinely expensive, storage and bandwidth and engineering all cost real money every month, and a company charging a modest subscription has to keep selling to new customers to keep the servers on for the old ones. A responsible founder in this space knows they might not make it, and the responsible thing they can do is exactly what several of these companies did: build an export, post a notice, keep the inbox open for a while. HereAfter is letting people email for their recordings. StoryFile found a buyer rather than simply going dark. By the ordinary standards of software, that is not villainy. That is a normal, even considerate, corporate death.
If a photo-sharing app I liked shut down with thirty days of export, I would grumble and move on. The files are replaceable, the stakes are low, and I understood from the start that I was renting a service, not buying a vault. Holding a struggling startup to the standard of a national archive would kill the category before it could do any good, and there is real good in it. The impulse to record a grandparent while you still can is one of the healthiest things technology has ever encouraged.
So the objection is correct about software in general. It is wrong about this specific product, and the reason it is wrong is the reason this whole essay exists.
A photo app that closes costs you files. A legacy company that closes costs a family the person, a second time.
Why is a legacy company's death different from a normal shutdown?
Because this category made a different promise, and it made it to people at their most trusting. Every other app sells you a service for now. A legacy company sells you permanence: record your father, preserve his stories, keep his voice so your children can hear it. That word, permanence, is the entire product. And the cruel structure of it is that the promise is only tested at the exact moment the customer can no longer act on it, after the person has died and the family goes looking for the voice they were told would always be there.
Researchers who study this have started to name the specific harm. Cambridge ethicists Tomasz Hollanek and Katarzyna Nowaczyk-Basińska have warned that AI recreations of the dead can "haunt" the bereaved, and that the industry building them needs safeguards for how these systems are retired, or fail to be retired, when a company changes course (University of Cambridge, 9 May 2024, published in Philosophy & Technology, DOI 10.1007/s13347-024-00744-w). Their scenarios cut both ways: a company that keeps a recreation running against a family's wishes, and a company that switches one off, or lets it decay, with no plan and no consent. Both are failures of governance, and governance is precisely the thing a shutdown strips away. When the company is gone, so is any promise about how the person you preserved will be treated.
This is where the ordinary tech-failure frame breaks. When Google Reader died, no one felt they had lost a relationship. When a legacy company dies, a widow can lose the only interactive record of her husband, the thing she described visiting the way you visit a grave. The grief researchers call the archive a continuing presence for a reason. You do not litigate for years, or drive to a server the way you drive to a cemetery, over something that means nothing. The meaning is real, which is exactly why the fragility is unforgivable. We have built a category that manufactures presence and then, on a balance sheet's timeline, can revoke it.
That is the turn. The question that opened this essay, can an AI really hold who someone was, is a philosophy-seminar question. The question that actually determines whether your family ever hears that voice again is a boring, unglamorous, commercial one: will this company outlast you, and if it does not, what did it build to hand your memories back? The interesting question has been getting all the attention. The boring question is the one with casualties.
What are you actually owed by a company selling permanence?
Owed is the right word, stronger than "should hope for." If a company takes money and a parent's voice on the explicit promise of preservation, the customer is owed specific, checkable things, and the fact that the industry has mostly not been held to them is the scandal hiding behind the more photogenic debate about realism. Here is the standard, stated plainly enough that you can hold any company to it, including mine.
You are owed real portability, not a rescue operation. Everything you put in should be exportable by you, on your schedule, in standard formats, at any time, not only after a farewell notice goes up and a support inbox becomes the last lifeline. The legal world already codified this shape of right: under Article 20 of the GDPR, a person can receive their personal data in a "structured, commonly used and machine-readable format" and move it elsewhere (GDPR Article 20). Treat that as the floor, not the ceiling. A recording of your mother should be as easy to walk out the door with as a bank statement.
You are owed consent that was actually given, by the living person, on the record. A legacy built from someone after they died, assembled from their leftover text and photos without their say, is not preservation. It is impersonation with good intentions. The person being preserved should have chosen it, chosen what goes in, and chosen what may happen later, while they were alive to mean it.
You are owed governance that survives death, written down and binding. Who can speak with the Persona, what the Persona may and may not do, when the Persona should fall silent: these cannot be left to whoever happens to own the servers next year. They have to be fixed by the person, in advance, in a form no later owner can quietly rewrite.
And you are owed the hardest thing of all: an honest answer to the company's own mortality. Any firm asking families to trust it for generations, while having no plan for its own end, is trading on a hope it cannot underwrite. "We'll be around forever" is not a plan. A plan is an export you already hold, terms that outlive the corporate entity, and a founder willing to say out loud that the company could fail and to design so that failure does not take your father with it.
Ask a legacy company one question before you trust it: what happens to my memories the day you no longer exist?
Notice what that standard is not. It is not "our AI is the most realistic." Realism is the marketing battle. Continuity is the real one, and it is won or lost in export policies, consent records, and governance terms that no one photographs for a launch. A company can lose the realism contest and still be the only one that keeps its promise, because it was the only one that planned for the day it would not be there to keep it in person.
The company selling permanence should answer for its own end
I run one of these companies, so this is written from inside the glass house, and the standard above is one I want used against Afterlife.ai®, not just by it. The failures in this essay are not competitors to gloat over. They are the strongest possible argument that this category has been building the wrong thing well: chasing believability while under-building the plumbing that decides whether a family ever gets the voice back.
So Afterlife AI™ is built around the boring question on purpose. Consent is structural rather than optional: a Persona is built by the living person represented, from what they chose to put in, with their agreement at each step, never assembled from someone after they are gone. That is why we will not offer to "recreate" a person from the outside after death, and never will; if the living person did not build their Persona, on purpose, that Persona does not exist. Executor Lock™ exists to answer the governance question the shutdowns expose: at the moment the person chooses, the record behind their Persona is frozen as a perfect snapshot, and every conversation after that draws only on it. A later party, the family, a new owner, or us, may govern access, but none of them can rewrite the record of who they were. Your Personas, conversations, memories, recipients and Trusted Contacts are yours to export as a structured data file on any ordinary day, so the export door is open in normal times and not only in an emergency. Voice is preserved with professional voice technology only where the living person has consented to it, including for playback after death, and that consent is locked and never changed later. We host in Australia, where a person's voice is treated as sensitive information under privacy law.
And here is what Afterlife AI™ cannot do, stated with the same plainness, because a company that only lists its promises has already broken the most important one. We cannot recover a recording trapped inside a competitor that has closed its export window. We cannot make any shutting-down company answer its support email. We cannot rebuild a person who never chose to be preserved, and we will not pretend the outputs are the person rather than a likeness the person built. And we will not overstate what walks out the door today: the file you can download now is a structured record of your written world, your Personas, conversations, memories, recipients and Trusted Contacts, and it does not yet package your original recordings and media, so carrying your voice and media out in standard formats is a commitment we hold ourselves to on this same standard, not a box we will pretend is already ticked. And we cannot prove today that we will still be here in fifty years, because no company on earth can. What we can do is design so that our survival is not the only thing standing between your family and what you preserved: you hold the written export now, the governance is fixed in advance, full media portability is a commitment on the record, and the standard is written down where you can hold us to all of it. If you are weighing what to preserve and where, our guide to who owns your digital afterlife covers the ownership question the checkboxes never ask, and the plans begin with a one-time free build so you can test the export before you ever trust the promise.
What should you check before trusting any legacy company?
You do not need to become an expert on bankruptcy law to protect a family archive. You need to ask a going concern the questions its marketing would rather you skipped, while you are alive and able to walk away if the answers are thin.
Can you export everything, right now, yourself? Log in today and try to download your data in a standard format. If the only path is emailing support, you are already relying on a rescue operation. Test the exit before you need it.
Where does the data actually live, and under whose law? A company hosting your voice under a serious privacy regime, and treating voice as sensitive information, has more constraints on abandoning it than one that has never said.
What is written down about death and shutdown? Look for consent that the living person gave, governance that survives them, and any statement at all about what happens to your files if the company closes or is sold. Silence there is the answer.
Who would own the archive if the company were acquired? StoryFile's memories changed hands in a bankruptcy court. Ask, before you upload, who inherits your father's voice if the founders sell.
The practical move underneath all four is the one thing entirely in your control: keep your own copy. If you have recordings in a company that is winding down right now, our walkthrough on saving your HereAfter recordings and the fuller picture of what HereAfter's shutdown means show exactly how to get them out while the window is open. A memory you hold in a standard file on your own drive cannot be shut down by anyone's board.
The reason to do this while you are alive is not morbid. It is the same reason anyone plans anything worth keeping: the person best placed to protect a voice is the person whose voice it is, deciding on purpose, before the choice belongs to a support queue or a probate court or a buyer you never met. The AI question, whether a machine can hold who you were, will keep being debated, and it is a good debate. But it was never the one that decides whether your grandchildren hear you. That one is decided by whether the thing holding you outlives you, and by what you insisted on being owed before you handed anything over.
Decide that part first. It is the only part that is still entirely yours.
Frequently asked questions
What happens to my recordings if a digital legacy company shuts down?
They do not automatically move anywhere safe. They remain on infrastructure the company has stopped maintaining, reachable only through whatever export it built while operating. Some companies post a notice and keep a support inbox open for a while, as HereAfter has done, but there is no regulator or automatic successor. The dependable protection is the copy you exported yourself, in a standard format, before any notice appeared, which is why you should test a company's export the day you sign up rather than the day you need it.
Can a company use my loved one's data after it is acquired or goes bankrupt?
Potentially, and that is the risk few people weigh before uploading. When StoryFile filed Chapter 11 in 2024, its assets, including its archives, were acquired by another company (PacerMonitor case 7:24-bk-22398; AV Interactive, 3 March 2025). Your data can change hands with the business, landing with owners you never chose. Before trusting any legacy company, ask what its terms say happens to your files in an acquisition or bankruptcy, and keep your own exported copy so the answer matters less.
Is my data safer with a big, established company?
Not necessarily. Size buys longevity, not permanence. Google is one of the most durable companies on earth, and it still deleted all consumer Google+ content on 2 April 2019 and closed Google Reader in 2013, each with only a short export window (TechCrunch, 2019 and 2013). A large company is less likely to vanish and entirely willing to retire a product that no longer fits its strategy. Judge any provider by its export rights and its written commitments, not by its market capitalization.
What is the difference between a company shutting down and my Persona being lost?
They are only the same thing if you never kept your own copy. A shutdown ends the service; it does not have to end your access to the underlying data, provided you exported it and provided the company built a real export in the first place. This is the entire argument for portability. The written record behind your Persona, your memories, conversations and account, is yours to download and hold as a structured file that outlives the company that hosted it, and original recordings and media should be held to the same portability standard, which is exactly the commitment to demand before you upload. So the right question about any provider is not "will you last forever" but "what can I walk out with today, and what has it committed to letting me carry out next."
How do I choose a digital legacy company that will not leave my family stranded?
Pick for continuity, not just realism. Confirm you can export everything yourself, right now, in a standard format. Check where the data is hosted and under what privacy law. Look for consent given by the living person and governance that survives their death, both written down. Ask directly who would own the archive if the company were sold. A provider that answers those plainly, and hands you your own copy without a fight, is worth more than the most convincing demo. Our guide to who owns your digital afterlife walks through the ownership questions in full.