What actually happens to a ChatGPT account when someone dies?

By default: nothing visible. The account stays active until the subscription lapses or OpenAI deactivates it for inactivity. Family members can request account deletion or data export by contacting OpenAI support with proof of death and legal standing, but there's no public 'legacy contact' system the way Apple, Google, and Meta have. The conversations sit there until someone with credentials logs in or the company is asked to act.

Can my family access my Claude or Gemini account?

Only if you've made provisions ahead of time. Google has Inactive Account Manager — you can name a trusted contact who gets access to specific services (including Gemini activity) after a defined period of inactivity. Anthropic does not currently offer a legacy contact program; access requires going through support with documentation. Microsoft has a similar process to Google for Copilot data. None of this is automatic.

Is there a way to make this easier for my family?

Three things help. First: a password manager with an emergency-access feature (1Password, Bitwarden) so a designated person can reach your accounts. Second: a written document explaining which AI accounts exist and what's in them — most family members don't know to look. Third: an export of the meaningful content (instructions, key conversations, distilled persona) stored somewhere accessible alongside other estate documents.

Is this morbid or sensible?

Sensible. The same conversation people now have about photos, voicemails, and journals. AI conversations are some of the most intimate records being created right now — therapy-adjacent thoughts, decisions about kids, plans nobody else heard. Treating them as part of an estate is the practical thing, not a dark one.

What does Konshus offer for this?

Konshus has a dedicated Crisis Handoff export (Counsel Pro tier) — a portable persona file plus selected conversations that a named contact can access if something happens to you. The vault itself is yours during your lifetime; the handoff is a clean, defined package designed for the moment after. We also have a broader Legacy export that lets you choose which atoms and conversations are included.

A guide · ~6 min read

What happens to your AI account when you die

Q: What about the AI persona itself — the version of me that lived inside the model?

That doesn't transfer to your family. The trained persona ChatGPT or Claude built across years of conversation is inside the provider's systems and isn't part of any export. The conversations are exportable; the model's internalization of who you were is not. If keeping a version of that persona accessible to people you love matters to you, it has to be exported and stored in something portable, not left inside the provider.

A question almost nobody asks until it's too late. ChatGPT, Claude, and Gemini hold years of conversation that, for many people, is more intimate than any journal they've kept. Here's what actually happens to that data when someone dies — and what you can do, today, to make it accessible (or not) the way you'd want.

By provider, in 2026

OpenAI (ChatGPT). No public legacy-contact program. Family members must contact support with documentation to request deletion or data export. Account remains active until the subscription lapses or inactivity triggers deactivation.
Anthropic (Claude). No public legacy program. Same support-and-documentation path. Account stays active until billing failure or manual deactivation.
Google (Gemini). Covered by Google's Inactive Account Manager — you can nominate a trusted contact in advance who gets access to specified data (including Gemini Apps Activity) after a chosen period of inactivity. The most mature legacy system of the major AI providers, and it requires you to set it up.
Microsoft (Copilot). Similar to Google — Microsoft account holders can designate a Next of Kin / deceased-user process, with data retrievable through support.
Apple (Apple Intelligence). Covered by Apple's Legacy Contact program (set up in Apple ID settings). The cleanest legacy story of any consumer tech company.

Notice the asymmetry: the big-platform players (Google, Microsoft, Apple) have legacy infrastructure because they've had it for decades. The AI-native players (OpenAI, Anthropic) haven't built it yet.

What to do today

Set up legacy contacts where they exist. Google Inactive Account Manager. Apple Legacy Contact. Microsoft Next of Kin. Ten minutes each. Do it.
Use a password manager with emergency access. 1Password's Emergency Kit and Bitwarden's emergency access both let a trusted contact reach your accounts after a defined waiting period.
Write a short index. A plain document listing which AI accounts exist, what's in each one, and what you'd want done with them. Most family members don't know to look at all.
Export the meaningful stuff. Custom Instructions, key conversations, important Projects. Keep them somewhere accessible alongside other estate documents — not only inside the AI provider that won't be reachable easily.
Decide what you'd want kept. Some people want the whole archive available to their family. Some want most of it deleted and a distilled "this is who I was" version preserved. Either is valid; making the call ahead of time is what makes it possible.

The thing that doesn't transfer

The trained persona ChatGPT or Claude built across years of conversation lives in the provider's systems. It isn't included in any export. The raw conversations are exportable. The model's internalization of who you were — how it knows what to assume about you, what tone to use, which topics matter — is not.

If preserving that part matters to you, it has to be built into something portable, outside any single AI. Otherwise it disappears the moment the account is closed or the model is deprecated.

Short answer

Set up legacy contacts where they exist (Google, Apple, Microsoft). Use a password manager's emergency access for the rest. Write a short index for your family of which AI accounts exist. Export the meaningful conversations and Custom Instructions to somewhere they can reach. The trained AI persona itself doesn't transfer — to preserve that, it has to live in a vault outside any single provider.

The portable answer

Konshus has a Crisis Handoff export — a portable persona and selected conversations that a named contact can access if something happens to you. There's also a broader Legacy export for choosing what to include. The vault is yours during your lifetime; the handoff is a defined, intentional package for the moment after, designed to be something a family member can actually open and use.

See also AI digital legacy for the broader frame and how to delete what AI knows about you for the other side of the same question.

Frequently Asked Questions

The version of you they kept. Backed up. Reachable.

Konshus builds a portable persona and a Crisis Handoff package so the version of you that lived inside ChatGPT, Claude, and Gemini isn't lost when the account is.

Meet Konshus