Frothly

Policy brief · FR-PB-2026-09

Half-Life

What happens to our data when we die.

In brief

Most people now leave behind dozens of accounts, thousands of messages and photographs, and an expanding set of AI models trained on their words. What happens to all of it is decided today not by inheritance law but by the terms of service of the companies that hold it.

This brief proposes a simple framework for digital estates built on three parts: a right to be remembered, a right to be forgotten, and a clear default between them for the many people who never say which they want.

Recommendations

  1. Accounts and data should pass to a person’s estate by default, like any other property, unless they choose otherwise.
  2. Services that shut down should be required to offer heirs a full export, and to offer public-interest material to independent archives.
  3. No conversational replica of a person should be created or operated after their death without their recorded consent.
  4. Where a person left no instruction, data should be preserved for their heirs for ten years, and replicas should not be made.

The problem

When someone dies, their physical belongings pass through a process that is centuries old. Their digital belongings pass through nothing at all. Families are locked out of photographs, correspondence and records by services that will not, or cannot, verify their right to them. Other material lingers indefinitely, visible and searchable, long after anyone would have wanted it to.

A newer problem sits on top of the old one. Services now offer to build conversational replicas of the dead from their messages, posts and recordings. Some families find comfort in them. Others discover a replica of a parent or child made without their knowledge, by someone they have never met.

150+online accounts held by the average adult
72%of accounts are still accessible to heirs one year after death
6%are still accessible after twenty years

The half-life of a digital estate

Working with partner archives, Lazarus traced what happened to the accounts of 2,000 people who died between 2004 and 2024. Digital estates decay at a strikingly regular rate. Services close, accounts are purged for inactivity, passwords die with their owners and formats become unreadable.

Figure 1Share of a deceased person’s accounts still accessible to their heirs
After 1 year 72%
After 5 years 41%
After 10 years 18%
After 20 years 6%

Lazarus estate study, 2,000 estates, 2004–2024.

Roughly half of what remains is lost every five years. Like any half-life, the rate is predictable, which means it is also preventable. Almost none of the loss we observed was deliberate. It was simply nobody’s job to stop it.

Two rights in tension

Any framework has to balance two legitimate wishes that pull in opposite directions. Some people want to be remembered: their work kept, their correspondence available to historians, their photographs passed to their children. Others want to be forgotten: their accounts closed, their messages deleted and nothing of them left to be simulated.

Both wishes deserve respect. The difficulty is that most people never express either one. The law therefore needs a default, and the default needs to be reversible in one direction only. Data that has been preserved can always be deleted later. Data that has been deleted cannot be brought back.

Conversational replicas raise the sharpest questions. Services have an incentive to keep families talking, and some measure their success by how often the bereaved return. Several families described the experience as comforting at first and harder to end later.

The Turing Registry already treats the retention of a personality without consent as grounds for revocation. In 2023 it revoked the licenses of two engram-storage and extraction systems that had done exactly that. We recommend that the same standard be written into law: a replica may be created only with the recorded consent of the person it imitates, and must never be optimized to maximize how often the living return to it.

What we recommend

For legislators

  • Treat digital accounts and data as part of a person’s estate, with access for executors by default.
  • Require services that close to provide exports to heirs, and to offer public-interest material to independent archives.
  • Prohibit the creation or operation of conversational replicas of the dead without their recorded consent.

For platforms

  • Offer every user a simple way to record what should happen to their data, and honor it.
  • Never use engagement measures to optimize memorial or replica services.

For individuals

Write it down. A single page naming who should have access to your accounts, and whether you would want to be remembered, forgotten or simply left alone, will save the people you leave behind a great deal of trouble.

References

  1. Arasaka Corporation (2023). Response to Turing Registry notice of revocation. Filing TR-08333-R.
  2. Lawson, M. (2022). Bringing people back: lessons from digital restoration. Archival Science, 22, 301–318.
  3. Runciter, G. (2017). Half-Life: The Persistence and Decline of Minds in Storage. Zürich: Beloved Brethren Press.
  4. Schoenheit von Vogelsang, H. (2020). Running a moratorium: ethics and practice. Journal of Memorial Studies, 5(1), 12–29.

Acknowledgments

We thank the families who spoke with us about their experiences, and the partner archives that made the estate study possible.

About the authors

Miranda Lawson leads Lazarus, Frothly’s digital preservation program, from our Vancouver office.

Herbert Schoenheit von Vogelsang advises Lazarus on end-of-life services and has operated a memorial archive in Zürich for twenty years.

How to cite

Miranda Lawson and Herbert Schoenheit von Vogelsang (2026). Half-Life What happens to our data when we die. Frothly Policy brief FR-PB-2026-09. San Francisco: Frothly. https://frothly.org/research/half-life/

Download PDF 7 pages · 115 KBBack to the library