There is a question the AI industry has been avoiding.

Not because it is complicated, though it is. Not because the technology is immature, because it is not. But because the answer is inconvenient for a lot of companies building in this space right now.

The question is this: whose voice is it?

When an AI platform clones the voice of someone who has died, using recordings they never consented to train a model with, who authorized that? And when a family member interacts with that voice, believing it represents their loved one, what exactly are they interacting with?

These are not hypothetical questions. They are happening now, across multiple platforms, and the industry has largely moved forward without answering them.

The Technology Moved Faster Than the Ethics

AI voice cloning has become genuinely impressive in a short period of time. Given a few minutes of audio, a well-trained model can reproduce someone's voice with enough accuracy to be emotionally convincing. The warmth, the cadence, the particular way someone pauses before they laugh. It is technically remarkable.

That technical capability is why the digital legacy space attracted so much attention so quickly. If you can recreate someone's voice, the possibility of preserving them in a meaningful way starts to feel real.

But the technology's speed created a problem. The ethical frameworks that should govern its use did not arrive at the same time. Most platforms launched into the space without answering the foundational question: does the person whose voice is being cloned get to decide whether that happens?

For many platforms, the answer has quietly been no.

What Unauthorized Voice Cloning Actually Looks Like

The most visible version of this problem is in entertainment. Deceased musicians, actors, and public figures have had their voices recreated without the consent of their estates, let alone their own consent given while alive. The legal battles are ongoing, the outcomes are unclear, and public opinion is genuinely divided.

But the digital legacy space has a quieter version of the same problem.

Some grief tech platforms are designed to help bereaved families interact with an AI representation of someone they have lost. The families upload recordings, home videos, voicemails, whatever they have. The platform trains a model on those materials and generates a voice and personality that sounds like the person.

The motivation is understandable. Grief is real, and the desire to hear someone's voice again is one of the most human responses to loss.

But the person being represented never agreed to this. They did not choose what recordings to use, what memories to include, what aspects of themselves to preserve or protect. They did not select who would have access or under what circumstances. They had no input at all.

What gets built is not their legacy. It is the family's interpretation of who they were, assembled from whatever happened to be recorded. Those are different things.

Most of the public conversation about posthumous voice cloning focuses on rights. Who owns the voice? Can the estate consent on behalf of the deceased? What does intellectual property law say?

These are important questions and the legal landscape is still forming around them. Several US states have passed or are considering legislation around digital likeness rights. The EU is developing frameworks under existing personality rights law. It is genuinely unsettled territory.

But there is a separate problem that gets less attention: unauthorized cloning produces an inaccurate representation.

A person's voice is not just acoustics. The way someone speaks is inseparable from how they think, what they value, and what they chose to say in any given context. When you clone a voice without the person's participation, you get the surface without the substance. You get a sound that is recognizable but a presence that is incomplete.

Understanding what a genuine AI digital legacy requires makes this clearer: the technology can capture a voice, but capturing identity requires the person's own input over time. No amount of posthumous audio processing closes that gap.

The result is that grief tech platforms built on unauthorized data are not giving families what they think they are getting. They are giving them something that sounds right but represents a person who never had the chance to say: yes, this is how I want to be remembered.

Where the Line Should Be

There is a meaningful distinction between two kinds of platforms in this space, and it matters more than most people realize when they are choosing where to build their legacy.

The first kind is built for grief. It is designed for people who have already lost someone and want to reconstruct a version of them from what remains. Whatever the intention, this requires working around consent because the person is no longer there to give it. The product is built from what the family has, not from what the person chose to leave.

The second kind is built for legacy. It is designed for people who are alive, thinking about the future, and want to participate in building what they will leave behind. The person is present, choosing what to include, how to present themselves, who will have access. The product is built from deliberate acts of preservation, not from whatever recordings happened to survive.

Most people searching for ways to talk to a lost loved one are looking for the second kind and finding the first. That gap is the central problem the digital legacy industry has not solved honestly.

EchoVault's Position on This

EchoVault only builds Echoes of living people.

That is not a marketing position. It is an architectural one. The platform is designed around the idea that an accurate, lasting representation of a person requires their own participation. Everything in the system, the voice, the personality, the memory architecture, the visual presence, flows from input that the person themselves provides.

The voice clone in an EchoVault Echo is built from recordings the person made deliberately. They chose what to say, how much to share, what tone to capture. The model is trained on their participation, not on data harvested from wherever it could be found.

The same principle applies to visual presence. EchoVault's video avatar system is built exclusively from the living person's own likeness, captured with their knowledge and consent. EchoVault does not build video avatars of people who are no longer here to approve them. There is no workflow in the platform for uploading someone else's recordings and generating an avatar on their behalf.

This is not because the technology could not be adapted to do that. It is because doing so would produce something inaccurate and ethically indefensible, and accuracy over time is the only thing that makes a digital legacy worth building.

The people who will interact with an Echo, the designated Echo Custodians who gain access when the time comes, deserve to know that what they are interacting with was built by the person it represents. Not reconstructed by grieving family members doing their best. Not assembled from whatever audio files could be found. Actually built, with care, over time, by someone who knew exactly what they were leaving behind.

What This Means for the Industry

The digital legacy space is growing fast and the ethical standards are still being written. Platforms that move quickly and ask questions later will capture early users. Some of them are already doing that.

But the category will eventually be judged by longevity and trust. A platform that creates an unauthorized representation of someone and gives it to their family is taking on a significant liability, legally and relationally. When the representation drifts, when it says something the person would never have said, when a grandchild asks a question thirty years from now and gets an answer that was never authorized, the harm is not abstract.

The platforms that survive the long arc of this industry will be the ones that treated consent as a foundation rather than a feature. Not because regulators will force them to, though that may come, but because accurate, consent-based representations are simply better products. They hold up. They remain trustworthy. They do not become something the family quietly stops using because it started to feel wrong.

Your voice is yours. What happens to it after you are gone should be a decision you made, not one that gets made for you.

That is the premise EchoVault was built on. It is also, we think, the only premise worth building on.

Frequently Asked Questions

Is it legal to clone someone's voice after they die? The legal answer varies significantly by jurisdiction and is still actively being litigated. Several US states have passed laws around digital likeness rights for deceased individuals, particularly in entertainment. Outside of those specific statutes, the legal framework is unsettled. Whether something is currently legal and whether it is ethical are, in this case, genuinely separate questions.

Can my family build an AI version of me from my old recordings without my permission? Technically, yes — the tools exist to do this. Whether any specific platform will help them do it depends on that platform's policies. EchoVault will not. The system requires the person being represented to be the one building their Echo. Old recordings uploaded by a third party are not a supported input.

What is the difference between grief tech and legacy tech? Grief tech is built for people who have already lost someone. Legacy tech is built for people who want to leave something behind while they are still alive. The distinction matters because consent is only possible in the second case. An accurate representation requires the person's participation, and participation requires them to be here.

Why doesn't EchoVault build video avatars of deceased people? Because a video avatar requires likeness consent that a deceased person cannot give. Visual presence is one of the most personal aspects of someone's identity. Building an avatar from old photos or videos without the person's authorization produces something that looks like them but was never approved by them. EchoVault's video avatar system is built from the living person's own deliberate recordings, nothing else.

What happens to my EchoVault Echo if I stop using the platform while I'm alive? Your Echo remains accessible to your designated Echo Custodians under the terms you set. What the platform will not do is continue building or expanding the representation based on third-party input. The Echo reflects what you built. That does not change when you are no longer here to add to it.

How do I start building my legacy before it is too late? The honest answer is that there is no urgency signal that tells you when to start. The richest Echoes are built gradually, over time, when memory is clear and voice can be captured at quality. The best time to start is while the thought is fresh. You can begin here →

EchoVault is a consent-based digital legacy platform. Only the living build their Echo — so the people they love always know exactly who they are talking to. Start building yours →