Meta's patent to simulate dead users: what the Facebook death patent does, and how to keep control of your AI likeness

Facebook death patent: the popular name for Meta's US patent 12,513,102 B2, an AI trained on one user's activity that keeps acting as that user when the user is absent or has died.

Persona: the AI version of you that you build on Afterlife.ai™ while you are alive, from your own memories, values and cloned voice.

Executor Lock™: Idy's patent-pending system of named executors, Trusted Contacts, a verified death report and a sealed snapshot of your Persona that can never be changed.

The Facebook death patent is US 12,513,102 B2, a patent Meta was granted on December 30, 2025 for an AI that learns from one user's activity and keeps acting as that user, including after the user dies. The official title is "Simulation of a user of a social networking system using a language model", the owner is Meta Platforms Technologies LLC, and the only inventor listed is Andrew Garrod Bosworth, Meta's chief technology officer (Google Patents). Meta filed the application on November 29, 2023. The abstract says a social networking system "simulates a user using a language model trained using training data generated from user interactions performed by that user."

The interactions the patent lists are "commenting on posts, liking content items, sharing content items, posting content items, sending messages to other users, going to an event, broadcasting messages on a wall, and so on." In plain words, the model studies what you do on the network and learns to do the same things the way you would. The patent also describes permissions in which the target user chooses which kinds of activity may be used to train the model, so the person being simulated can limit what the model learns from. The table quotes the passages that made headlines exactly.

Read together, the five passages describe a system that writes your posts, answers your messages and speaks on calls with your voice and face after you have stopped. Those permissions cover what goes into the model while you are alive, but the passages about death name nobody, such as an executor or a family member, who decides whether the bot keeps running once you are gone. The rest of this page shows you how to settle that question for yourself.

Yes, the Facebook death patent is real, granted and active, and anyone can read the full text on the public record. The number and the grant date appear on Google Patents and in the US Patent and Trademark Office's full-text PDF, and the American Bar Association's intellectual property newsletter analyzed the claims in "Posting Posthumously: Analyzing Meta's User-Simulation Patent". The viral coverage began on February 16, 2026, when Dexerto reported that Meta had patented an AI that takes over a dead person's account to keep posting and chatting. Business Standard and MediaPost followed on February 17, Vice on February 18, and Fortune on March 3, 2026.

A granted US patent lets its owner stop others from using the claimed invention in the United States, generally for 20 years from filing, which puts this one at about 2043. A grant does not oblige the owner to build anything. It does prove that Meta described the method in detail and that a patent examiner accepted the claims as new, which is why the patent matters even if Meta never ships the product. Any company that wants to build the same system in the United States now has to work around Meta's claims or license them.

Meta says it is not, and the company has announced no product that simulates dead users. A Meta spokesperson told Business Insider, "We have no plans to move forward with this example," as MediaPost reported on February 17, 2026 and Fortune repeated on March 3, 2026. So if you are asking whether Meta is going to use AI to bring back dead Facebook users, the company's answer on the record is no. Large technology companies file thousands of patents a year, and many never become products. The more useful question is what Meta does build. Since 2023 Meta has launched AI characters played by celebrities, let creators build AI versions of themselves on Instagram, and begun building an AI version of its own chief executive, and that trend, more than the patent, is the reason to put your own instructions in writing now.

Project Lazarus is the name an anonymous poster gave to Meta's dead-user idea in February 2023, nine months before Meta filed the patent, and Meta has never confirmed the name as an internal project. On February 23, 2023, someone claiming to be a Meta insider and a developer on "Project Lazarus" posted on 4chan that the company was building an AI that could "take over a deceased persons social media accounts" and keep posting as if the person were alive. Vision Times reported the post four days later and judged the claim "almost certainly 4Chan disinformation" at the time (Vision Times, February 27, 2023).

The name came back when the patent was reported. MediaPost ran its February 17, 2026 story under the headline "Meta's 'Project Lazarus' Would Resurrect The Dead", reported that Meta had called the idea Project Lazarus three years earlier, and noted that Gab's chief executive, Andrew Torba, had reposted the original image from an unknown source (MediaPost). The same report recalled a 2023 interview on Lex Fridman's podcast in which Mark Zuckerberg described wanting to create virtual avatars of people who have died. The patent itself never uses the name Project Lazarus, so the accurate summary is that the name comes from an anonymous 2023 post, and the method is now a granted Meta patent.

The patent gives Meta's own reason, which is that a user's silence hurts the people who follow that user, and death makes the silence permanent. The filing says that "the impact on the users is much more severe and permanent if that user is deceased and can never return to the social networking platform." The cyberpsychologist Elaine Kasket told Fortune that this reasoning "suggests that user death is like an engagement problem" (Fortune, March 3, 2026). The scale explains the concern, since Meta's apps were used by 3.60 billion people a day on average in June 2026 (Meta, July 29, 2026), and every one of those people will eventually stop posting. The dated record below lists every public step Meta has taken on death and on AI versions of real people since 2009, and we re-check the record every month.

At least 1.4 billion Facebook users will die before 2100, and the number could pass 4.9 billion if Facebook keeps adding users, according to Carl Öhman and David Watson of the Oxford Internet Institute, writing in Big Data & Society in 2019 (DOI 10.1177/2053951719842540). The 1.4 billion floor assumes Facebook stopped gaining users in 2018, and in that scenario the dead could outnumber the living on the platform by 2070, according to the University of Oxford's release reported by Phys.org on April 27, 2019. Most of those profiles will belong to users outside the West, and in the growth scenario Nigeria alone accounts for more than 6% of the total. The authors warned that a purely commercial approach to the data of the dead carries ethical and political risks, and their numbers show that the question reaches every ordinary account. Our guide to what a griefbot is carries the original projection chart.

No, Facebook cannot recreate dead people today, and no Meta product offers an AI version of someone who has died. If you are asking whether Facebook or Meta AI can recreate your mother, your father or other dead relatives, the answer in October 2026 is no, and Meta has said it has no plans to build the system its patent describes. What Facebook offers is a memorialized profile, introduced in 2009, and a legacy contact, introduced on February 12, 2015 (Meta Newsroom). A memorialized profile shows "Remembering" next to the person's name, keeps the posts and photos the person shared, and never posts on its own.

If someone you love has died, the most valuable step you can take this week is to save what already exists. Download the photos and videos you were tagged in, save the voice notes in your own message threads with that person, and ask relatives to do the same. If you are the legacy contact and the person allowed downloads, Facebook lets you download a copy of what the person shared, and for the wider question of AI and loss, read what a deadbot is.

Your Facebook account is either memorialized or permanently deleted when you die, depending on the choice you made, or on a request from your family if you made none. A memorialized account shows "Remembering" next to your name, nobody can log in to the account, and your posts and photos stay visible to the audience you shared them with. You can choose one legacy contact to look after a memorialized account, and you must be 18 or over to choose one, or you can tell Facebook to delete your account permanently after your death (Facebook Help Center). Our full guide to what happens to your Facebook account when you die covers every setting step by step.

A Facebook legacy contact can look after your memorialized profile, but cannot log in as you, read your messages or speak for you. The table lists the powers exactly as Facebook's Help Center described them when we checked on October 2, 2026 (Facebook Help Center). None of these powers mentions an AI version of you, because the settings were designed for a profile page long before models were trained on profiles. To give one person authority over your accounts, files and AI likeness together, read what a digital executor does, and for Meta's messaging app, read what happens to WhatsApp when you die.

Your Meta AI memories sit in a separate record from your posts, and the legacy contact and memorialization settings above were written for your profile rather than for what you told the assistant. Meta said on January 27, 2025 that Meta AI would remember certain details people share with the assistant in one-to-one chats on WhatsApp and Messenger (Meta). Since December 16, 2025, Meta has also used people's interactions with its AI features to personalize the content and ads they see across its apps in most regions, as it announced on October 1, 2025 (Meta). What you tell the assistant therefore becomes part of the record Meta holds about you, and our guide to what happens to your Meta AI memories when you die covers every setting, the deletion options and what your family can and cannot reach. Other assistants offer families no more, and SpaceXAI can now delete a Grok account inactive for more than 120 days, as what happens to your Grok conversations when you die explains.

You can object to Meta using your public posts for AI training if you live in the European Union or the United Kingdom, and in the United States the practical options are narrower. Meta announced on April 14, 2025 that it would train its AI on public posts and comments shared by adults in the EU, along with people's questions to Meta AI, and that every EU user would get a link to a form to object at any time (Meta). Ireland's Data Protection Commission cleared the plan, and training on EU data began on May 27, 2025 (The Register, May 22, 2025). Meta says it does not train on the content of private messages with friends and family unless someone in the chat shares those messages with its AI, and that it does not use public data from EU accounts of people under 18.

MIT Technology Review reported that users in the United States and other countries without national data privacy laws have no foolproof way to stop Meta using their data to train AI (MIT Technology Review, June 14, 2024). The steps that work everywhere are to keep the posts you care about private, to review and delete what Meta AI remembers about you, which Meta says you can do at any time, and to decide now what happens to your accounts after you die. The patent shows why the last step matters, because a model trained on your activity could keep acting as you after your death, and the only reliable answer is a written instruction backed by a person you named.

No, the Meta and Microsoft patents are two different documents from two companies, granted five years apart, and the main difference is where the training data comes from. Microsoft's US 10,853,717 B2, "Creating a conversational chat bot of a specific person", was granted on December 1, 2020 from a priority date of April 11, 2017 (Google Patents). Microsoft's chatbot learns from a person's "social data", which the patent defines to include images, voice data, social media posts, electronic messages and written letters, and the patent adds that a 2D or 3D model of the person may be generated from images and video. Meta's model learns from activity inside one social network and adds simulated audio and video calls. Microsoft said it had no plans to build the product, and Tim O'Brien, who managed AI programs at Microsoft, called the idea "disturbing", as Fortune recalled in March 2026.

Meta launched 28 AI characters on September 27, 2023, several of them played by celebrities, and retired the celebrity-styled versions in less than a year. The characters ran on Meta's Llama 2 model and included Kendall Jenner as "Billie", Tom Brady as "Bru", Charli D'Amelio as "Coco", Snoop Dogg as "Dungeon Master", Paris Hilton as "Amber" and a character played by MrBeast (TechCrunch, September 27, 2023). TechCrunch reported that the characters were built in cooperation with the celebrities but ran entirely on AI. By July 2024 Meta had quietly retired them, after the bots fell flat with users, according to The Information, and had moved its effort to AI Studio, where creators build AI versions of themselves (Social Media Today, July 31, 2024).

The lesson for anyone thinking about an AI of themselves is about authorship. Each celebrity lent a face to an invented character with a different name and personality, so fans who came for Kendall Jenner met "Billie", and the experiment ended within a year. An AI that answers in your name earns trust when the answers come from you, in your own words, which is why an Afterlife.ai™ Persona is built only from your own memories and conversations. Meta's characters also appear in our record of what happened when HereAfter AI shut down, beside other AI services that closed or changed course.

Creators could make an AI version of themselves on Instagram through Meta's AI Studio from July 29, 2024, and in August 2026 Meta stopped anyone creating new AI characters or editing existing ones. At launch, Meta said a creator could set up an AI "as an extension of themselves that can quickly answer common DM questions and story replies", with responses labeled as AI (Meta, July 29, 2024). An update dated August 10, 2026 says people "will no longer be able to create new AI characters or edit existing ones" as Meta focuses on experiences powered by Muse Spark, and Meta launched its Muse personal agent in the United States on September 8, 2026 (TechNode Global, September 9, 2026).

AI Studio was built to answer fans inside one app, with no executor and no plan for the creator's death, and a creator who set one up in 2025 cannot edit that AI today. An Afterlife.ai™ Twin answers on doors you own, which are your website, any app you connect through the API and the AI assistants people already use, reached over MCP. You read every conversation, pause any door at any time, and decide in writing what happens to your twin when you die.

Yes, Meta is building an AI version of Mark Zuckerberg, a Meta AI twin of its own chief executive trained on his "mannerisms, tone, public statements and thoughts on company strategy", so that Meta's employees, about 79,000 at the time of the report, can talk with the AI when he is unavailable. The Financial Times first reported the project, and Entrepreneur summarized the report on April 14, 2026, adding that Meta believes the experiment could be replicated by influencers and creators (Entrepreneur). BBC Future reported in August 2026 that Meta has been building the twin so employees feel more connected to their boss (BBC Future, August 12, 2026). The company that patented a way to simulate dead users is now cloning its own chief executive, and AI versions of real people are arriving whether or not anyone has written rules for them.

The AI Zuckerberg is authorized by the person and trained on his own words, and an Afterlife.ai™ Twin gives you that same authorized version of yourself. Your twin answers in the first person, under your name and photo, from the Persona you built, and visitors who speak hear your own cloned voice. Afterlife Pro, which includes the twin, costs $59.99 a month or $599 a year (every price on this page is in US dollars), and your twin keeps answering after Executor Lock™ where you switch that on. For the full product, read about the Afterlife.ai™ Twin.

Whether any company could legally make an AI of you after you die depends on where you live, what the AI is used for and what you agreed to in the platform's terms, and in the United States, control of your likeness after death comes from state law. Indiana and Oklahoma protect the right of publicity for 100 years after death, California for 70 years, and Tennessee's right never expires as long as the use continues (Wikipedia, Personality rights). On December 11, 2025, New York's governor signed a law requiring consent from heirs or executors before anyone uses a dead person's name, image or likeness for commercial purposes (New York Governor's Office), and the federal NO FAKES Act, which would create a national voice and likeness right, is still a bill. Platform rules add another layer, and X, for example, lets families only deactivate a public account, as what happens to your X (Twitter) account when you die explains.

Outside the United States, the European Union has required since August 2, 2026 that people be told when they are talking to an AI, Denmark's broadly agreed bill would give everyone rights over AI imitations of their face, voice and body for 50 years after death, and China's draft rules on digital humans would ban AI clones of people made from their personal information without consent. Most of these laws aim at commercial use, labeling and consent from the living, so for an ordinary user inside a platform's own service, the platform's terms and your own written instructions decide most of the outcome. For the detail behind each row, read digital replica laws by state, the NO FAKES Act explained, deepfake and AI likeness laws by country, what Article 50 of the EU AI Act means for you and who owns your AI twin after you die.

You stop anyone making an AI of you after you die by putting your wishes on every platform that holds your data, writing them into your will, naming a digital executor, and building the one AI of you that you control. The seven steps take about two and a half hours in total, and each one closes a different gap. Start with the platforms, because their own settings usually govern their own accounts. Step seven carries the most weight, because a Persona under Executor Lock™ gives people the version of you that you do want, and once your executors accept the Executor Lock™ Agreement, the product enforces your choices after your death. For the identity question in general, read who controls your identity after you die.

Meta's afterlife patent reconstructs a person from the activity the person left behind, while Afterlife.ai™ starts with the living person, who builds a Persona on purpose and decides in writing what happens later. The patent's model infers a personality from posts, likes and messages that were never written for that purpose, and the network that runs the bot decides when the bot runs. On Afterlife.ai™ you author your Persona memory by memory, and your named executors take over under rules you set while alive. Consent against reconstruction is the whole difference, and the table sets the two designs side by side.

The voice row shows the gap clearly. The patent generates speech and video of the user for calls, while an Afterlife.ai™ Persona speaks in a voice you recorded and approved yourself, and Persona Video makes video messages in your own face and voice from one 30-second recording of you. Nothing in your Persona comes from scraping your social media, and our guide to a digital twin after death covers the rest.

An executor matters because an AI of you can outlive you, and someone has to hold authority over that AI with your rules in hand. Without a named person, the company that hosts the AI decides, and Meta's patent shows where a company's own logic can lead. Executor Lock™ names that person in advance and follows the same steps for every Persona.

1. Name your people. You name your executors and Trusted Contacts, and each one accepts the Executor Lock™ Agreement while you are alive.

2. Report and check. After your death, an executor or Trusted Contact uploads evidence such as a death certificate, and automated analysis and a human review check the evidence.

3. Cooling-off. Seven days pass in which every executor can dispute the report, and you can cancel a false report with proof of life.

4. Lock and release. Your Persona becomes a sealed snapshot of every memory as the memory stood that day, nothing pruned, and your release rules for photos, memories and live conversation take effect.

After the lock, nobody can edit a memory or change who you are, and our Terms make the lock permanent and irreversible (section 8.3). New short-term memories and big life events can still reach your Persona, so a grandchild can share good news while your Persona stays exactly who you were, as how Executor Lock™ works explains. Your executor can still invite the people you named, hand the role to someone else or ask Idy to delete your Persona.

Idy, the company behind Afterlife.ai™, built this governance layer before building the twin, because an AI of a person is safe to put to work only once control is settled. Our first years went into who controls your likeness while you live, who decides about your likeness when you die, who may see which memories, and how consent is recorded so the software obeys the consent. That work is covered by 78 patents pending, including verifying a death before control of a Persona changes hands, handing authority to a named executor, sealing a memory record that becomes permanent at the handover, and keeping a Persona the same person when the underlying AI models change. Idy is also seeking international protection under the Patent Cooperation Treaty, and in 2025 Idy filed more standard patent applications than any Australian organization except Aristocrat, ahead of CSIRO, ResMed and the University of Melbourne, according to the national patent office's Australian IP Report 2026.

Because that foundation came first, one Idy twin can answer as you while you live and pass to the people you named when you die. Meta's own experiments show the narrower designs on offer elsewhere, since the celebrity characters were invented personalities wearing famous faces and AI Studio's creator AIs were built to answer fans inside one app. The same holds for video-avatar tools made for marketing clips and for creator chat clones made to answer an audience, and none of them asks you to name the person who takes over when you die. For the wider category, read an AI chatbot of a dead person and what happens when a digital legacy company shuts down.

An Afterlife.ai™ Persona starts free, and the Afterlife.ai™ Twin comes with Afterlife Pro at $59.99 a month or $599 a year. The free build holds 25 memories, needs no card and includes Executor Lock™ setup, so the most important decisions cost nothing. The yearly Pro price saves $120.88 against twelve monthly payments, and twins are for people 18 and over. The usual path is to start free, name your executor on the first day, and add memories over a few evenings. For every plan in detail, read every Afterlife.ai™ plan and price.

You should act now because the patent is granted, Meta is cloning its own chief executive, and any AI of you made later will follow whatever rules exist on that day. The laws in the table above are still arriving state by state and country by country, and none of them was written with your accounts, your messages and your voice in mind. The only rules that are certainly yours are the ones you write while you are alive, and the platform settings take minutes while the will clause and the named executor take an afternoon. A Persona you build this month, with your executors named, gives your family the version of you that you chose, and the free build costs nothing.

1. Build your Persona free, no card

2. Name your executor

3. Switch on your Afterlife.ai™ Twin

We read both patents on Google Patents, Facebook's Help Center, Meta's announcements and quarterly results, the 2023 and 2026 reporting linked above, the Öhman and Watson study and the laws in our state and country guides. Every outside fact links to its source and date, and we re-check the status table every month.

• who owns your AI twin after you die

• what a digital executor does

• what a griefbot is

• deepfake and AI likeness laws by country

• the Afterlife.ai™ Twin