A recently surfaced patent granted to Meta in 2023 described a system that could use artificial intelligence to simulate a user’s social media presence — even after death. The idea, which the company now says it has no plans to pursue, has reignited debate over digital identity, grief and the expanding ambitions of large language models.
A Patent That Stirred Unease
It is not difficult to imagine why the concept unsettled critics.
In 2023, Meta was granted a patent outlining how a large language model could “simulate” a user’s activity on a social networking platform. The document described using past posts, messages and other digital traces to generate new content in a person’s voice when the user was absent — whether temporarily or permanently. The patent explicitly noted that such simulation could apply “when the user takes a long break or if the user is deceased.”
Under the proposal, a digital replica could continue interacting with others through comments, likes and even direct messages. The patent stated 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,” suggesting that the system was designed to account for the emotional weight of such scenarios.
The primary author listed on the patent is Meta’s chief technology officer, Andrew Bosworth. While the company has since distanced itself from the idea, the patent’s existence underscored how far technology companies were willing to explore new uses for large language models during a period of intense experimentation.
Certified Cyber Crime Investigator Course Launched by Centre for Police Technology
Engagement, Data and Business Incentives
The proposal emerged at a moment when social media platforms were confronting shifting user behavior and declining engagement in certain markets. Facebook, once defined by lively personal updates, has increasingly become a repository of dormant accounts, automated birthday reminders and algorithmically curated content. At the same time, critics argue that feeds have grown saturated with what some describe as “AI slop” — low-quality, machine-generated posts optimized for clicks.
As engagement fluctuates, advertising — Meta’s core business — remains closely tied to user activity. More interactions mean more data, and more data strengthens the feedback loop that fuels targeted advertising and future AI systems.
“It’s more engagement, more content, more data — more data for the current and the future AI,” Edina Harbinja, a law professor at the University of Birmingham, told Business Insider when the patent became public.
She said she could see the business incentive behind such a move, adding that she was curious how and whether the company would implement the innovation.
For companies racing to expand the capabilities of large language models, few datasets are as rich as the billions of personal posts accumulated over decades on social platforms. The patent illustrated a willingness to treat those archives not only as historical records but as material for generative systems.
Grief, Memory and Ethical Boundaries
The notion of training AI on the posts of deceased users drew swift concern from scholars and ethicists.
“One of the tasks of grief is to face the actual loss,” Joseph Davis, a sociology professor at the University of Virginia, told Business Insider. “Let the dead be dead.”
The debate over digital afterlives predates generative AI. As social media went mainstream, questions arose about what should happen to dormant accounts: Should they remain untouched as memorials? Should family members gain access? Or should platforms close them altogether?
In recent years, technology has complicated those questions. Examples have surfaced of AI systems trained to emulate deceased individuals — from a grandmother “resurrected” as an AI model for her funeral to startups in the so-called “grief tech” sector offering tools that allow loved ones to train models on images, recordings and messages of the departed.
Such projects promise comfort through simulated interaction. But critics warn that digital replicas may blur boundaries between remembrance and replication, potentially altering the grieving process.
Even Mark Zuckerberg appeared to acknowledge the tension. In a 2023 interview with the podcaster Lex Fridman, he suggested that virtual avatars might one day take over accounts of deceased people, allowing interactions to continue.
“If someone has lost a loved one and is grieving, there may be ways in which being able to interact or relive certain memories could be helpful,” he said at the time.
He added, however, that such uses could also become unhealthy and would require further study.
“We have, you know, a fair amount of experience with how to handle death and identity and people’s digital content through social media already, unfortunately,” he said.
A Shift in Tone
In the years since, public attitudes toward generative AI have shifted. As AI-generated content has proliferated across platforms, enthusiasm has been tempered by concerns about authenticity, misinformation and over-automation.
The conversation around the patent reflects that change. What may once have been viewed as a speculative extension of digital memorialization now appears, to some observers, as emblematic of a period when companies were “throwing everything at the wall” to discover novel use cases for large language models.
Meta now says it does not intend to move forward with the example outlined in the patent. “We have no plans to move forward with this example,” a spokesperson told Business Insider.
Still, the document remains a snapshot of a transitional era in technology — a moment when the boundaries between memory, identity and machine learning were being actively redrawn. As platforms continue to grapple with vast archives of human expression, the question of what becomes of those digital traces after death remains unresolved, even if one particular path has been set aside.
