Meta has been granted a patent for artificial intelligence that could allow a person’s social media account to keep interacting after they die.
The system would use a large language model trained on a user’s historical activity to simulate how that person behaves online, potentially responding to posts and messages as if they were still present.
Filed in 2023 and approved in late December, the patent lists Chief Technology Officer Andrew Bosworth as its primary author. It describes training AI on posts, comments, likes, and other activities to replicate a user’s tone and interaction patterns. Meta told Business Insider it has “no plans to move forward” with the concept.
How the system would work
According to the filing, the model could simulate a user when they are “absent from the social networking system,” including when the user is deceased. Once trained, the AI could respond to content, reply to direct messages, and maintain ongoing engagement based on prior behavior.
The patent also references the possibility of simulating video or audio calls, extending the system beyond text.
Meta presents the idea as a continuity tool, designed to reduce disruption when someone stops posting. In practice, that would mean keeping the account active rather than leaving it dormant.
More than a memorial page
The patent makes clear that this would not be a static archive. The goal is participation. The AI would generate new interactions modeled on the individual’s communication style. That distinction shifts the concept from remembrance to replication. A memorial page reflects a life. A simulated account continues the conversation.
The filing also notes potential use for creators or influencers who step away but want to maintain engagement. Applied after death, however, the same mechanism enters more sensitive territory.
The unanswered operational questions
Although Meta says it is not pursuing the feature, the patent leaves key issues unresolved.
Consent is not detailed. The document does not explain how users would authorize posthumous simulation or what role families might play. It also does not address whether other users would be clearly informed that they are interacting with an AI model.
Data boundaries are another open question. Training on years of personal activity raises questions about scope and limits. And while a model can approximate someone’s style, it can also generate responses the person never wrote, creating the risk of divergence from the real individual.
A signal about AI’s expanding reach
There is no product roadmap attached to this patent. But filings often reveal where companies are experimenting technically, even if nothing ships.
In this case, the direction is clear: generative AI is being explored as a way to simulate identity itself. Whether that capability becomes a feature remains uncertain. What is not uncertain is that the technical possibility is now on the table.
For users, the question is simple and unsettled: if AI can keep your account active after you’re gone, should it?
Lawmakers in the European Parliament are now working without native AI features on their devices as cybersecurity concerns take center stage.