Griefbot Estate Planning: Consent, Privacy, and Family Boundaries
A griefbot can sound like a comfort tool until the family has to decide whether one should exist.
The idea is simple: an AI system is trained on messages, recordings, photos, videos, posts, or other personal material so it can simulate a person who has died. It may appear as a chatbot, voice clone, video avatar, or memorial companion. Some people call these tools griefbots, deadbots, postmortem avatars, or AI memorial chatbots.
The estate planning question is not whether the technology is impressive. It is whether your family should be allowed to make an interactive version of you, what material they may use, who gets to talk to it, and when it should stop.
That question should not be left to guesswork. A griefbot can touch private messages, family memories, voice recordings, identity rights, children, and grief itself. It can comfort one relative while upsetting another. It can preserve a sense of connection, but it can also create confusion, dependency, conflict, or the feeling that a person is being made to speak after death without permission.
Griefbot estate planning turns that emotional problem into a documented decision. It gives survivors a rule to follow before they are approached by a service, a relative, a business partner, or a well-meaning friend who says, "We can bring their voice back."
What a Griefbot Is
For planning purposes, define a griefbot broadly.
It can be any AI system that imitates a deceased person in an interactive or newly generated way. That includes a text chatbot that replies like the person, a voice clone that says new words, a video avatar that appears to speak, or a memorial tool that generates new letters, advice, jokes, or birthday messages.
This is different from keeping an archive. Saving real photos, messages, and recordings preserves what the person actually created. A griefbot creates new output. Even if it is based on real material, the system is still generating words or sounds the person did not personally choose.
That difference is why consent matters. A family photo album usually does not need the same rules as an AI voice that can say, "I forgive you," "sell the house," or "I would have wanted this." The more a system appears to speak as the person, the more careful the plan should be.
Why This Belongs in a Digital Estate Plan
Most digital estate plans cover passwords, devices, social accounts, photos, cloud storage, and financial records. Griefbots add a new layer: posthumous simulation.
Researchers in Philosophy & Technology describe the digital afterlife industry as involving several affected groups. There is the person whose data is used, the person who possesses or uploads that data, and the person who interacts with the resulting system. Those roles may overlap, but they are not always the same.
That matters because the person with the data may not be the person who should decide. A sibling may hold years of text messages. A spouse may have voice notes. A business partner may have recordings. A child may want an AI conversation. None of that automatically proves the deceased person consented to being simulated.
The U.S. Copyright Office has also warned that realistic digital replicas can harm private individuals, not only famous people. The risk is not limited to celebrities, performers, or politicians. Ordinary families have voices, faces, messages, and histories that can be copied too.
Start With a Clear Consent Rule
Your instruction should begin with one of three positions.
The first is no consent. This means no one may create, maintain, publish, license, or interact with an AI simulation of you after death. Your family may preserve real materials, but they may not use them to generate new conversations, voice clips, or avatar performances.
The second is limited private consent. This allows a narrow memorial use, such as a private, clearly labeled chatbot for named adults in the family. This option must define what data may be used, who may interact with it, whether children are excluded, whether outputs can be saved, and when the system must be deleted.
The third is project-specific consent. This may fit a creator, teacher, founder, performer, or public figure who wants a particular archive, documentary, course, or memorial project. It should name the project and should not become a blank permission slip for every future tool.
If you are uncertain, choose a restrictive default. Survivors can still preserve memories through real photos, recordings, and letters. What is difficult to undo is a public simulation that spreads before the family has agreed on boundaries.
Decide What Data Can Be Used
A griefbot is only as private as the material used to create it.
List allowed and prohibited sources. Possible source material includes texts, emails, chat logs, social posts, photos, videos, voice notes, meeting recordings, journals, medical records, family documents, and business files.
Private messages deserve special caution. A message sent to one person was not necessarily permission for a whole family, company, or vendor to train a simulation. Emails may include other people's secrets. Journals may contain thoughts never meant for anyone else. Voice notes may be emotionally intimate. Medical, legal, and financial records should generally be excluded.
A useful instruction might say: "Public talks and published writing may be used for an estate-approved archive. Private texts, emails, journals, family videos, medical records, and legal documents may not be used to train or prompt an interactive simulation."
That kind of detail helps the decision-maker say no without turning every memory into an argument.
Name the Decision-Maker
Do not let every relative become an equal approver.
Name one person who can interpret your wishes and approve, refuse, pause, delete, or retire any griefbot. This may be your executor, digital executor, spouse, adult child, sibling, trustee, close friend, or another trusted person. Name a backup too.
Then define conflicts. A person who wants to sell access, promote a project, settle a family argument, or keep a business running may not be neutral. You can say that anyone with a financial conflict cannot approve a griefbot without another named person's consent.
The decision-maker should also have authority to request deletion, revoke access, report impersonation, and stop a service if it becomes distressing or unsafe.
Protect the People Who Interact With It
Consent is not only about the deceased person. It is also about the living.
A surviving spouse may want to talk to the system every night. One child may find it comforting. Another may feel disturbed. A parent may want a young child to hear an AI version of a deceased relative. A relative may feel pressured to participate because everyone else says it is healing.
Your plan should make interaction optional. No family member should be required to use, watch, hear, fund, or maintain a griefbot. If you allow one, say who may use it and who may not. Consider excluding minors unless a parent, guardian, therapist, or other appropriate adult has carefully evaluated the situation.
The Philosophy & Technology paper recommends adult-only access for these services. Even if a family chooses a different rule, children deserve special care because they may not understand what is real, simulated, archived, or invented.
Require Clear Disclosure
A griefbot should never pretend to be the person.
Require labels such as "AI simulation," "generated voice," or "memorial chatbot based on archived material." The label should appear inside the interface, on any shared output, and before a person interacts with the system for the first time.
The EU AI Act includes transparency obligations for certain AI-generated or manipulated content, including deepfakes, subject to exceptions. Your family instruction can use the same basic principle even outside the EU: people should know when they are interacting with a simulation.
Disclosure also protects memory. A generated answer may sound emotionally true, but it is not the same as a real statement. The system can invent, smooth over contradictions, or reflect the hopes of the person who uploaded the material. Families should not treat its output as evidence of the deceased person's actual wishes.
Ban High-Risk Uses
Even if you allow a private memorial chatbot, you can prohibit risky uses.
Consider banning griefbot outputs for legal, medical, financial, political, romantic, sexual, employment, business, fundraising, advertising, and public endorsement contexts. A simulated person should not appear to sign off on estate decisions, advise children, settle family disputes, endorse products, solicit donations, or continue a relationship in ways the living person never approved.
Voice cloning adds another risk: impersonation. The FTC has warned about AI-enabled voice cloning harms, including fraud. Families should treat voice output as especially sensitive and should avoid publishing clean voice clips that could be reused by scammers.
If a voice feature is allowed at all, require short access windows, private use, no downloads, and clear labels on every clip.
Plan for Retirement
A griefbot should have an ending.
Researchers have recommended procedures for retiring deadbots. That may sound strange, but it is practical. A service may become emotionally unhealthy, technically insecure, expensive, inaccurate, or unwanted. A child may grow older and feel differently. A family may discover that the tool is not helping.
Your instruction should say when the system must be reviewed and when it must be deleted. For example: review after 30 days, again after one year, and delete unless the named decision-maker renews access. Require export of real source materials before deletion, if appropriate, but do not preserve generated conversations by default.
Retirement is not disrespect. It can be the boundary that keeps a memorial tool from becoming a permanent obligation.
Put the Instruction Where Survivors Can Find It
A griefbot instruction belongs with your digital estate plan, digital legacy letter, estate binder, password manager emergency notes, or attorney-held documents.
Keep the language plain. Say what is allowed, what is prohibited, who decides, what data may be used, who may interact, how the system must be labeled, and when it must end.
Then tell your decision-maker where the instruction lives. The best boundary in the world will not help if no one finds it until after the service has already been created.
A Simple Clause to Adapt
Here is a plain-language starting point:
"I do not consent to an AI chatbot, voice clone, video avatar, or other interactive simulation of me after death unless this document specifically allows it. My family may preserve real photos, messages, recordings, and writing, but they may not use private material to generate new speech or conversation in my name. If I allow any limited memorial AI use, it must be private, clearly labeled as artificial, approved by my named decision-maker, unavailable to children unless separately approved, and deleted when the decision-maker determines it is no longer helpful."
Adjust that clause with an attorney if it needs to coordinate with a will, trust, publicity rights, intellectual property, creative contracts, or business succession plan.
The Bottom Line
A griefbot is not just another account setting. It is a decision about identity, privacy, memory, and the emotional lives of survivors.
Some families may eventually use these tools with care. Others will decide that real archives are enough. Both choices can be respectful. The important thing is that the choice should come from the person being simulated, not from a vendor's promise or a grieving family's guess.
Write the boundary now. Your survivors should inherit your memories, not the burden of deciding whether an artificial version of you should speak.