Meta is developing a deeply personalized AI assistant intended to serve its more than 3 billion users, according to people familiar with the effort who spoke to the Financial Times. The project is part of the company's broader push into agentic tools - software that can act autonomously on behalf of users.
At the center of the work is an advanced digital assistant powered by Meta's new Muse Spark AI model. One source familiar with the program said a subset of Meta employees are testing the assistant internally, indicating the project remains in an experimental stage within the company.
People involved describe the effort as aiming for capabilities similar to OpenClaw - a platform that lets users design AI bots, called agents, which can carry out tasks autonomously. Meta's stated goal is to enable users to create and rely on agents that handle a range of everyday activities without continuous manual input.
Crucially, the company intends to give users the option to share highly sensitive personal information with these assistants if they choose to do so. That includes health and financial data, according to one person familiar with the plans. The intention to permit such data sharing underscores the degree of personalization Meta is targeting, but it also raises questions about user acceptance and trust.
Some observers cited in the reporting expressed skepticism about whether consumers will be comfortable providing that level of private information to an AI assistant, even if the sharing is optional. The concern highlights privacy and adoption uncertainties surrounding product designs that rely on intimate personal data.
For now, details remain limited to accounts from people familiar with the project. The testing phase among staff suggests the company is still refining the assistant's capabilities and user controls before any broader release.
Context and next steps
Meta's effort to produce an agentic assistant using Muse Spark reflects a push to embed more autonomous AI into consumer-facing products. How broadly and quickly such a product might be rolled out is not clear from the available reporting; the company appears to be in an internal testing phase.