Meta Secures AI Patent To Simulate Absent Accounts, Even After Death

Meta Platforms was granted US patent for a system that retrains a pretrained language model on a person’s historical social-network activity to generate actions and responses “on behalf of the user,” explicitly including when the user is deceased.

The patent’s core claim is behavioral continuity. A social networking system “simulates a user using a language model” trained on training data generated from that user’s interactions, and the model “may be used for simulating the user” during prolonged absence “or if the user is deceased.”

The system receives a pretrained language model, retrains it using “user specific training data” derived from prior platform interactions, then deploys it so a bot can generate predicted interactions in response to content items relevant to the user, including feed items.

One described control is user-granted permissions that allow or disallow specific interaction types for training, for example allowing comments but excluding private messages from training data.

The patent also contemplates multiple models per user anchored to different life stages, with separate language models trained up to points-in-time corresponding to ages (example ages given include 20, 25, and 30).

Business Insider reported Meta said it is not proceeding, quoting a spokesperson: “We have no plans to move forward with this example,” while emphasizing that granted patents can be disclosures rather than product commitments.

The same report frames the patent as an extension beyond static post-mortem presence toward automated engagement, and situates it alongside prior “grief tech” concepts, including Microsoft’s earlier patenting of a chatbot that could imitate real people using social data.

Information for this briefing was found via Business Insider and the sources mentioned. The author has no securities or affiliations related to this organization. Not a recommendation to buy or sell. Always do additional research and consult a professional before purchasing a security. The author holds no licenses.