One Sentence in Your Will Decides Whether an AI of You Is a Legacy or a Violation

Yes. A will can carry a clause that gives named people permission to use your recordings, writing and data to build a representation of you after your death, and that names who may speak with the result and who may not. No statute I have read forbids that sentence, and several treat your name, voice and likeness as property that passes under a will.

A will is a written, signed and witnessed statement of what you want done with what you own after you die, carried out by the executor you appoint. The clause we are discussing asks that old instrument to carry something the law has only recently begun to treat as yours: the sound of your voice and the shape of your face.

I told EL MUNDO's Ricardo F. Colmenero, in the interview that ran on August 30, 2026, that there are things Afterlife AI will not do even if doing them would pay. Recreating a person who has died is one of them, whatever recordings the family holds. The single exception I could imagine is a person who left explicit written consent, for example in their will, that their data may be used to build a personality like this. That is consent. Today it almost never exists.

So the sentence can be written, and on its own the sentence is a wish with a signature: more than most people leave, less than a family will need.

Consent written down beats consent guessed at, every time.

A will controls your property and gives instructions to your executor. It controls your data only where a law lets the executor reach that data, and your likeness only where a law treats likeness as property. Three separate questions, and most people assume the first answer covers all three.

An executor is the person your will appoints to carry out its instructions and administer your estate. Executors are good at bank accounts and houses, and less good at a phone that locks itself, an account held under terms written in another country, and a company that never heard of your will.

In the United States the gap has a partial fix. The Revised Uniform Fiduciary Access to Digital Assets Act is a model law, completed by the Uniform Law Commission in 2015, that extends a fiduciary's power over tangible property to a person's digital assets. Executors, trustees and agents under a power of attorney can manage files, domains and virtual currency, but the Act, in the Commission's words, "restricts a fiduciary's access to electronic communications such as email, text messages, and social media accounts unless the original user consented in a will, trust, power of attorney, or other record." The priority order matters: an online tool the platform offers comes first, a will or trust second, the platform's terms of service third. The RUFADAA adoption map shows which states have a version.

That is what a will can reach: your executor, and through RUFADAA, your accounts and, with express consent, their contents. What a will cannot reach is a platform's retention policy, data already deleted under that policy, and any company outside the jurisdiction the will was made in.

A will speaks to your executor. Everyone else is listening through a wall.

In the United States it depends on the state, and in the two states with the strongest post-mortem rights it depends on whether your identity had commercial value when you died. Europe is taking the disclosure route rather than the ownership route, and Denmark is attempting ownership for everyone. None of it was written with an ordinary family's Persona in mind.

A right of publicity is the legal right to control commercial use of your name, voice, signature, photograph or likeness. In some states that right survives death and can be passed on; in most it protects careers rather than people.

California is the model everyone copies. Civil Code section 3344.1 protects a "deceased personality," defined as a person whose name, voice, signature, photograph or likeness "has commercial value at the time of that person's death, or because of that person's death." The right runs for 70 years, passes by contract, trust or testamentary instrument, and cannot be enforced for damages until the successor registers with the Secretary of State. Since AB 1836 the section also reaches a digital replica, "a computer-generated, highly realistic electronic representation that is readily identifiable as the voice or visual likeness of an individual." Read the definition again: an accountant from Fresno whose face had no commercial value is not a deceased personality, and her children have no claim under 3344.1 at all.

New York's section 50-f, signed on November 30, 2020 and in force 180 days later, runs the same way for 40 years and adds a rule for deceased performers: their digital replica may not be used "in an audiovisual work, sound recording, or for the live performance of a musical work" without consent. The law follows the money. Tennessee's ELVIS Act, signed on March 21, 2024, added voice for every individual, "regardless of whether the sound contains the actual voice or a simulation of the voice of the individual."

Federally there is a bill. The NO FAKES Act of 2026, S.4591, was introduced on May 20, 2026 and placed on the Senate calendar on June 24. It has passed neither chamber. If it becomes law it would create a licensable federal property right in every individual's voice and visual likeness, with liability for platforms that knowingly host an unauthorised replica. What the NO FAKES Act would actually cover reads the bill rather than the press release.

Europe chose labels over ownership. Article 50 of the EU AI Act applies from August 2, 2026 and requires whoever deploys artificially generated or manipulated image, audio or video of a real person to disclose that "the content has been artificially generated or manipulated." That is a transparency duty, not a right you can leave to your children; what Article 50 means for you sets out the limits. Denmark is trying the other thing: a draft amendment to the Copyright Act, published on July 7, 2025, would give every natural person a consent right over realistic digitally generated imitations of their appearance and voice, lasting 50 years after death. We could not confirm final adoption at the time of writing.

The law protects the famous dead. Everyone else has a will and a wish.

Because a will is read late, in public, by an executor who may not hold your data, against companies that never agreed to it. Each of those four is a hole. Together they are the reason a clause without a build behind it protects almost nothing.

Probate is the court process that proves a will and supervises the executor's administration of the estate. In California the courts' own guidance says formal probate "typically takes 9 to 18 months and can sometimes take even longer," and the process opens with a petition, a notice to heirs and a notice published in a newspaper. Public by design, slow by design. A phone with a 90-day inactivity policy does not wait for a hearing.

Late is the first hole; gone is the second. The recordings a clause refers to live on devices that lock, in accounts that close, under retention policies nobody in the family has read. The third hole is the platform: under RUFADAA the platform's own online tool outranks your will, and where there is no tool and the will is silent, the terms of service govern. A company in another country is reading its own policy, not your probate file.

The fourth hole is the one that bothers me most, because it is not about access. A will can say yes. A will cannot say how. When Ricardo F. Colmenero asked me where a real person ends and an AI character begins, I gave him the example I use with my own team. "I like Spanish football" is a fact you said. "My favourite player is X and I think we will win" is a conclusion the model drew from it. A sentence in a will gives permission for the first and, without meaning to, licenses the second. It cannot say which memories count, which recording is the real one, who may ask what, and whether the personality may keep changing after you are gone.

That is the difference between consent and governance, and it is why the strong version of this idea is not a better clause. The strong version is consent recorded while you are alive, attached to a Persona you built yourself, and locked. Executor Lock is the process that freezes a Persona as a perfect snapshot at verified death: every memory kept, nothing added to the personality, no retraining, no drift. How Executor Lock works sets out the mechanics, and who controls your identity after you die covers the wider question.

A will says yes or no. A lock says exactly what yes means.

Start building your legacy Free build, 50 memories, no card required.

An AI clause in a will should say three things: that you consent, what may be used, and who may and may not have access. Everything below is illustration, not legal advice. Wills are governed by the law of the place you live, and a clause that works in New South Wales may fail in New York. Take these words to an estate lawyer in your jurisdiction and let them redraft.

A permission clause, for a person who has built a Persona:

During my lifetime I created and consented to a digital representation of myself (my Persona) held in my account with Afterlife AI (Idy Pty Ltd, Sydney). I direct my executor to preserve that account and its release settings, to take no step to alter, retrain or extend the personality recorded in that account, and to allow access only to the people I named within the account. I do not consent to any other person or company creating a representation of my voice, face or personality from my data, recordings or communications.

A permission clause, for a person who has built nothing yet:

I consent to my recordings, writing and data being used after my death by [named person] to create a digital representation of my voice and personality, for the private use of my family only, provided it is built from my own words and recordings, is never used for commercial purposes, and is not altered or extended to say things I did not say.

A prohibition clause:

I do not consent to the creation, after my death, of any digital representation of my voice, face or personality by any person or company, and I direct my executor to refuse any request to provide my data, recordings or communications for that purpose.

Four notes on all three. Name the account and the person, because an executor cannot honour a direction to "my family." Say what may not happen, because the inference problem above is the thing most families will not anticipate. Put the same words in a signed, dated letter kept with the will and, where the platform allows, inside the account itself, because the will is the last document anyone reads. And remember that a clause is consent, not enforcement: in California and New York your heirs' ability to sue depends on registration and commercial value. The clause tells the people who love you what you wanted. It does not make a stranger stop.

Name the account, name the person, and say what may not happen.

Say yes or no while you are alive, out loud, to the people who will otherwise have to guess. Then write the same answer down where they will find it. The conversation is the consent; the paper is the proof.

If a daughter asks, and she will ask in a car or at a sink rather than at a table, the answer that helps her is specific. Yes, and I will build it myself so you are not building it from voicemails. Or: yes, but only from what I record on purpose, only for you and your brother, and nobody sells it. Or: no, and I would rather you kept the videos. Any of those is a gift. The vague version, we'll see, is the one that ends in the message we receive most weeks.

If you want the opposite, say so plainly and put a prohibition clause in the will. Then close the gaps a clause cannot: set the online tools that outrank your will, a Legacy Contact or an inactive-account setting, so your accounts are deleted or handed over on your terms. A written no gives your family standing to refuse a well-meaning relative, and gives any company that respects consent a reason to decline.

There is a second reason to decide, and it is the one I gave EL MUNDO. A Persona you validated yourself is a defence: an authorised version makes an unauthorised version harder to justify and easier to refuse. A prohibition does the same work in the other direction. Silence leaves everything to whoever reaches your data first.

Decide out loud. Then write it down where they will find it.

Today, mostly, nobody is checking. That is the objection at full strength and I concede it. Most companies in this industry do not ask whether the person consented, a will binds your executor and not a startup in another country, the NO FAKES Act is a bill, the state laws protect commercial value, and platforms follow their own terms. A clause on its own is a wish with a witness.

Here is the turn. A clause is weak when it has to create something. A clause is strong when it only has to point. If the Persona already exists, built by you from your own words and recordings, with consent stored against your account, with the people who may speak with your Persona named while you were alive, and with Executor Lock waiting at verified death, then the will has one job: to say, this exists, I made it, honour its settings. An executor can do that. A court can read that. A company that respects consent can act on that without a hearing.

That is where my own company enters, late, and the claim is narrow. Afterlife AI, built by Idy Pty Ltd in Sydney, does not recreate anyone who did not consent while alive, because a company in this industry is defined by what it refuses. The one exception is explicit written consent from the person themselves, and even then the build is held to the same rules: their own material, locked, never retrained. The buyer is alive. The free build is 50 memories, no card, and never expires. Release runs through Executor Lock: nomination by name, verification of the evidence, a dispute window, then a locked snapshot. Data is hosted in Australia. Trust is our foundation sets out how a reader can test that rather than take my word for it.

Three things this does not do. A locked Persona does not stop a stranger assembling a fake from your public audio, and no product ever will. Your heirs gain no claim under 3344.1 if your face had no commercial value. And a Persona is a version of you, never you.

A sentence in your will is the difference between a legacy and a violation. A sentence that points at something you already built is the difference between a wish and a fact.

Write the clause. Then build the thing it points at.

Can my family make an AI of me after I die?

Only if you said so while you were alive, in writing. Without that, no company that respects consent will build one, and Afterlife AI will not. With that consent, a will clause or a signed letter naming who may use your recordings, for what, and with what limits is enough to let a family act. The better version is to build your Persona yourself now, from your own words, and let the will point at it, so your family is opening something you finished rather than guessing from a folder of voice notes.

What is an AI clause in a will?

An AI clause in a will is a sentence or short paragraph that records your consent, or your refusal, to a digital representation of your voice, face or personality being created or used after your death. A good one names the person or account it refers to, says what material may be used, forbids alteration or commercial use, and lists who may have access. It binds your executor. It does not bind platforms or companies outside your jurisdiction, so treat it as consent, not enforcement, and have an estate lawyer draft it for the place you live.

Does the NO FAKES Act cover a dead person's voice?

The NO FAKES Act is not law. S.4591 was introduced in the Senate on May 20, 2026, reported by the Judiciary Committee and placed on the Senate calendar on June 24, 2026, and had passed neither chamber when this page was written. As drafted it would create a federal, licensable property right in every individual's voice and visual likeness, with liability for platforms that knowingly host unauthorised replicas. Until Congress passes it and the President signs it, a dead person's voice is protected only by state law, and in most states only where the voice had commercial value.

Can I forbid an AI of me in my will?

Yes. A prohibition clause states that you do not consent to any digital representation of your voice, face or personality after your death and directs your executor to refuse requests for your data for that purpose. Pair it with the settings that outrank a will, such as a Legacy Contact or an inactive-account rule that deletes or hands over your accounts on your terms, and tell your family out loud. A written no gives them standing to refuse a relative and gives any company that respects consent a reason to decline.

Does Afterlife AI build a Persona of someone who has already died?

No. We receive that request most weeks, often from someone holding recordings and wedding videos, and we decline every time, because the person cannot consent and cannot correct a word. The single exception is explicit written consent left by the person themselves, for example in their will, that their data may be used for this purpose. Even then the build follows the same rules as everyone else's: their own material only, locked at verified death, never retrained. If you have arrived here grieving, grief support lists people who can help.

• The lines we will not cross: every refusal, with the reason and what it costs us.

• How Executor Lock works: the mechanism the will only has to point at.

• Who controls your identity after you die: the wider question this page sits under.

• Digital replica laws state by state: the American picture beyond California, New York and Tennessee.

• What the NO FAKES Act would actually cover: the bill, not the press release.

• Accessing a deceased person's iPhone: the practical side of digital assets after death, and the Legacy Contact your will cannot replace.

• Afterlife AI vs ChatGPT: why a legacy has to stop learning, which is the clause the will points at.

• How does Afterlife AI work?: the Lock, the training rule and the release, in plain terms.

• Afterlife AI pricing: what the will can point at for free, and what paying keeps in place.

• California Legislative Information, Civil Code section 3344.1

• New York State Senate, Civil Rights Law section 50-f

• Manatt, Tennessee's ELVIS Act Expands Publicity Rights for Individuals' Voices

• GovInfo, Bill Status for S.4591, NO FAKES Act of 2026

• Holland and Knight, Senate Committee Advances Bill to Protect Name, Image, Likeness and Voice

• Uniform Law Commission, Fiduciary Access to Digital Assets Act, Revised (2015)

• Uniform Law Commission fact sheet on RUFADAA, hosted by the Alaska Bar Association

• EU AI Act, Article 50

• European Parliamentary Research Service, The Danish approach to copyright and deepfakes, January 2026782611_EN.pdf)

• Plesner, Personal identity meets copyright: Denmark moves to regulate deepfakes in the Copyright Act

• Global Law Experts, Denmark Deepfake Law 2026

• Danish Ministry of Culture, Bred aftale om deepfakes giver alle ret til egen krop og egen stemme

• California Courts Self-Help Guide, Overview of formal probate

• EL MUNDO, Chris Williams, CEO de Afterlife: "En cinco u ocho años podrás hablar con el holograma de un familiar fallecido en la mesa de Navidad", by Ricardo F. Colmenero, August 30, 2026

Chris Williams is the founder and CEO of Afterlife AI, a consumer brand of Idy Pty Ltd in Sydney, and the architect of Executor Lock. He has been interviewed on consent-first digital legacy by EL MUNDO, Channel 10 News, The Daily Telegraph and ABC radio. He is a father of four. Before Idy he founded Natural Solar, an Australian solar energy company. Nothing on this page is legal advice; a will should be drafted with an estate lawyer in your own jurisdiction.