Is the Facebook death patent real?

Yes. It is US Patent 12,513,102 B2, "Simulation of a user of a social networking system using a language model," assigned to Meta Platforms and granted on 30 December 2025. It was filed in November 2023 and lists several inventors, including Meta CTO Andrew Bosworth. The document genuinely describes training a language model on one user's own data and notes it could simulate a user who is absent or deceased.

Is Meta actually building this?

There is no public evidence that it is. When asked, Meta reportedly said it has no plans to move forward with this example. It is important to remember that a patent is not a product. Big tech companies file many patents defensively, and most are never built. A granted patent shows what a company has the right to attempt, not what it has decided to ship.

Is this the same as the Microsoft dead-chatbot patent?

No, and they are often confused. The Microsoft patent is US Patent 10,853,717 B2, "Creating a conversational chat bot of a specific person," granted in December 2020 to Microsoft Technology Licensing. The Facebook one is the 2025 Meta patent (12,513,102 B2). They are two separate filings, two different companies, about five years apart. Both are real; neither is a shipping product.

Could Meta simulate me after I die using my Facebook account?

The patent describes a method that could, in principle, use a user's own posts, messages, likes and voice data to simulate them. But describing a method in a patent is not the same as deploying it, and Meta reports no plans to build this example. Whether such a thing could ever run, and under what consent and legal conditions, is a separate and unsettled question the patent does not answer.

Why is consent the central concern?

Because the unsettling scenario is one where a simulation of you is assembled from your data without your active, informed consent and switched on after you can no longer object or correct it. The technology itself is not new. What people fear is a version of them existing that they never agreed to. That is why a consent-first approach, where you choose to build and govern your own digital self, is the meaningful difference.

How is Afterlife AI different from what this patent describes?

With Afterlife AI you build your own Persona while you are alive, with your consent at every step, rather than a company reconstructing you from a dormant account. Executor Lock lets you decide what is accessible after death, and that choice is locked and never changed afterwards, with consent covering posthumous access. Afterlife AI is an Australian company with Australian-hosted data. You can start free, with a one-time build budget of 60 memories and 100 conversations that never expires.

Yes, Meta really patented a way to simulate you after you die

The "Facebook death patent" went viral for a reason: it is real, it is on the public record, and it is about training an AI on your own posts to keep "you" talking once you are gone. Here is what it actually says, what the headlines got wrong, and why the only question that matters is consent.

Reviewed: 28 June 2026 · Explainer by Afterlife AI

In late 2025 a single patent filing did something patents almost never do: it went viral. "Facebook patents tech to make you post after you die," the headlines screamed. People were unsettled, and fairly so. The idea of a company quietly building a version of you that keeps talking after your funeral hits a nerve that few technologies ever reach.

So we sat down and read the actual document, plus the reputable coverage around it. The short version: the patent is real, the viral framing is mostly accurate, and a few of the loudest claims are wrong. Most importantly, the thing everyone is actually afraid of has a name, and it is not "AI." It is consent.

This page is published by Afterlife AI. We build consent-first digital legacy tools, so we have a point of view here. We have tried to keep the facts straight regardless, and to clearly separate what the patent says from what we think about it.

What the patent actually says

The document people are talking about is US Patent 12,513,102 B2, titled "Simulation of a user of a social networking system using a language model." The key facts on the public record:

Assignee: Meta Platforms (the company behind Facebook, Instagram and WhatsApp).
Inventors: the filing lists several names, including Andrew Bosworth, Meta's Chief Technology Officer.
Filed: November 2023. Granted: 30 December 2025.

What it describes is, in plain terms, this: take a general language model, then retrain it on one specific person's own data, the posts, comments, likes, private messages and voice data tied to that single account. The result is a model meant to respond the way that person would.

The line that lit the internet on fire comes from the patent itself. It says the language model "may be used for simulating the user when the user is absent from the social networking system, for example, when the user takes a long break or if the user is deceased." The filing also references simulated audio or video interactions, not just text.

That is the genuinely striking part, and it is true. A patent that openly contemplates an AI standing in for a dead person, built from that person's own account, on the world's largest social network.

Why it went viral (and what headlines got wrong)

The drama is real. But a few popular claims overshot, so here is the honest correction.

"Meta is building this." No evidence of that. When asked, Meta reportedly said it has no plans to move forward with this example. That matters, because of one fact that almost every viral post skipped:

A patent is not a product. Large tech companies file thousands of patents defensively, to stake out ideas and block competitors, and the overwhelming majority are never shipped. A granted patent tells you what a company has the *right* to attempt. It tells you nothing about whether it ever will.

"It will start posting as dead people tomorrow." There is no announced product, no rollout, no timeline. What exists is a legal document describing a method.

"They can do this to anyone." The patent describes a *technique*. Whether it could ever be applied, and under what consent and legal conditions, is a separate question that the patent does not, and cannot, settle.

None of this makes the story fake. It makes it precise. The patent is real and unsettling on its own terms. It just is not a confession of a shipping feature.

Wait, is this the Microsoft one? (clearing up the confusion)

If this story feels familiar, you may be thinking of a different patent. They get mixed up constantly, so let us separate them cleanly.

The Microsoft one: US Patent 10,853,717 B2, "Creating a conversational chat bot of a specific person," assigned to Microsoft Technology Licensing and granted back in December 2020. It also went viral, also described building a chatbot of a specific person from their social data, images, voice and messages, and even mentioned a possible 2D or 3D model. Microsoft is on record as not having plans to build it either.
The Facebook one: the 2025 Meta patent described above (12,513,102 B2).

They are two different patents, two different companies, about five years apart. When you see "they patented bringing back the dead," check the number. The Microsoft filing is from 2020. The Meta filing is the new one. Both are real; neither is a product.

The real issue: consent

Strip away the headlines and one question is left standing. Not "can a machine imitate a person?" That is already possible. The real question is: who decided?

The thing that genuinely worries people about a patent like this is the scenario where a simulation of you is assembled from your data without your active, informed consent, and switched on after you can no longer object. A version of you, speaking in your style, that *you* never agreed to and can never correct.

That is a legitimate concern, and it is the right thing to focus on. The technology is not the villain. The absence of consent is.

This is exactly the line we think matters, and it is the line we built our own product around.

The consent-first alternative: Afterlife AI

Afterlife AI™ is built on the opposite premise from the worry above. The point is not that a company quietly reconstructs you. The point is that you build your own Persona, yourself, while you are alive, with your consent at every step.

Here is how that differs in practice:

You are the author. You create your Persona from memories and conversations you choose to share. Nothing is scraped together behind your back. There is no "dead account taken over" scenario, because the whole thing only exists because you chose to build it.
Executor Lock™ governs what happens later. You decide what is accessible after you are gone, and that choice is locked. Once set, it is not changed after death. Your consent explicitly covers posthumous access, so the people you trust experience only what you intended, nothing more.
A consent-based voice, made with your permission. If you choose to preserve your voice, it is consent-based voice preservation of yourself while you are alive. That consent is locked at Executor Lock and never altered afterwards. The voice is created free for everyone; the listening experience is the paid part, and family inherits the time you have paid for. Nothing autoplays in a moment of grief; a family member always chooses to tap.
Australian company, Australian-hosted. Afterlife AI is an Australian company, and your data is Australian-hosted. Your voice is treated as sensitive information under Australian privacy law.

You can start building for free: a one-time build budget of 60 memories and 100 conversations, no card, no countdown, and your free build does not expire. If you want more, the public plans are simply Legacy at $14.99/mo and Eternal at $29.99/mo. Family inherits the time you have paid for.

The Facebook death patent asks a frightening question: what if a version of you outlives you, and you never agreed to it? Our answer is to make the agreement the entire foundation. You decide. You build it. You lock it. It is yours.

Frequently asked questions

The questions below cover the points people search for most after seeing the viral coverage. As always, we have tried to keep the patent facts separate from our own opinion.

Sources

Create your Persona See how it works Contact