4 Minutes
Imagine scrolling through your feed and seeing a reply from someone who passed away last year. Creepy? Intriguing? Both. Meta quietly secured a patent in December 2024 for a system that uses large language models to mimic user activity on social networks—even after the user is gone.
The filing, first submitted in 2023 under Andrew Bosworth's name, describes a “digital clone” that can simulate long absences from a platform. According to reports, the system would study a person's past posts, messages, and interaction patterns to generate future comments, likes, and even conversational replies that sound familiar to friends and followers.
Why would a company build something like this? One explicit use case in the document targets creators and influencers who rely on continuous engagement. They sometimes take extended breaks but don’t want the relationship with their audience to evaporate. The proposed bot acts as a digital proxy, keeping the persona alive in that narrow, commercial sense.
But the patent goes further. It mentions the possibility of the bot operating after a person’s death: responding to messages, liking posts, and simulating voice or video calls based on historical data so interactions feel natural. The language in the filing acknowledges a sharper emotional impact when the person being simulated has died, a frank nod to the psychological and social consequences of such a tool.
Meta's public line is cautious. A company spokesperson told reporters the idea is not part of the current product roadmap and active development has been paused. Ownership of the patent remains with Meta, though, leaving the door open for future licensing or revival. That combination—shelved today, protected for tomorrow—raises its own set of questions about intent and timeline.
This is not uncharted territory. Microsoft explored similar territory in 2021 with a patent aimed at chatbots that imitate people. That effort was later shelved after executives admitted it felt unsettling. The difference now is the sophistication of modern large language models and richer social data, which make simulated conversation and behavioral mimicry far more convincing—and potentially more ethically fraught.
There are clear technical appeals. Brands and creators prize persistent engagement. Automation that maintains voice and timing could preserve follower metrics and revenue streams. But what about consent, dignity, and the emotional fallout for friends and family who interact with a convincing replica? What legal frameworks govern a digital shadow that outlives its owner?
Regulators and ethicists will have to wrestle with scenarios that blur the line between legacy tools and deceptive impersonation. Does prior consent allow a platform to keep a persona active? Who controls the archive that trains such a model, and how transparent must the system be about its non-human nature? Answers will matter as much as the technology itself.
Meta's patent is a reminder that innovation often precedes consensus. The company has paused development, but the idea—digital continuity through AI—will resurface. It already has. And as models grow more adept at mirroring human quirks, the debate will shift from whether such systems can be built to how society wants them used.
For now, the patent sits like a loaded blueprint: technically feasible, ethically combustible, and waiting for a moment when business incentives, legal clarity, and public tolerance align.
Comments
Marius
Wow didn't expect that... getting replies from a 'digital version' of someone who's died would be so unsettling. For creators maybe useful, but for families, yikes. if that's real then...
mechbyte
Wait.. Meta wants bots to reply as dead people? is this even true? Feels illegal and gross, plus who controls consent? not cool, sounds like a privacy nightmare
Leave a Comment