In an investigation published October 6, 2026, TIME reported that Meta Muse builds profiles of users and people mentioned in their communications, including people who do not use Muse. The reported profiles may draw together relationship history, shared interests and conversations.
What TIME reported about Muse contact profiles
The reported scope reaches beyond the person using the agent: profiles may concern people who appear in that user’s communications, including non-users. That makes the distinction between an assistant remembering details about its user and building context about other people especially relevant.
NeoTeo previously covered reported instructions for Muse contact profiles; this follow-up looks at TIME’s separate report about profiles connected to users’ communications.
What Meta says Muse remembers and how it handles access
Meta says Muse can remember information about other people that users choose to share, then use it to personalize assistance. One example is remembering a friend’s dietary restrictions when planning an event.
Meta also says each user’s Muse runs in a dedicated cloud virtual machine. A separate agent called Sentinel must approve internet activity, and Muse asks for permission before sensitive actions such as sending an email or making a purchase. Users choose which connected apps Muse can access and can change permissions or disconnect a service.
Meta says users can opt out of using their interactions to train Meta AI models. The company also says Muse conversations and virtual-machine data are not shared with its advertising systems, and that Muse cannot see users’ passwords or payment methods.
What a request to forget means for the original message
Meta says users can ask Muse to forget specific information it has learned. TIME reported that a request may leave the original message containing that information in the chat. The learned memory and the message where it appeared are distinct: asking Muse to forget does not necessarily remove that source message.
