Meta’s Muse AI assistant may process details about people who do not use it when a user connects services such as email or a calendar, ZDNET reports. The practical wrinkle is that your information could enter an AI conversation through somebody else’s account.
Meta AI Watch analysis
What happened
ZDNET says names of meeting attendees and event details could be included when Muse users ask the assistant about their calendars. The same concern applies to messages and documents a user asks Muse to handle. The article quotes Meta saying that information about non-users may be included in requests and responses, and may be used or shared under Muse’s privacy policy.
Meta told ZDNET that such information is not stored in a way that allows it to reasonably identify which information belongs to a particular person. The report also says users can turn off the use of their interactions to improve Meta’s AI in Settings and Data controls. That setting is in the user’s hands, not those of the people whose details may appear in the material they share.
Why it matters
AI assistants can make other people’s information part of a task even when those people never opened the product or agreed to use it. A calendar entry, email thread or uploaded document can contain more than the account holder’s own details.
The distinction matters: Meta’s position, as reported by ZDNET, is that the information is not stored in a way that reasonably identifies a particular person. That does not mean the information cannot be included in a request or response. ZDNET also quotes privacy expert Calli Schroeder arguing that protections for people who do not consent are inadequate.
Our read
The useful rule is a social one with a technical edge: before connecting an assistant to your inbox or calendar, consider whose information comes along for the ride. Turning off AI training may limit one use of your interactions, but it does not prevent other people’s details from appearing in what you ask the assistant to process. Privacy settings are not a substitute for asking before sharing someone else’s sensitive information. A modest ask, perhaps, in an age of assistants eager to read everything.
What to watch
- Whether Meta adds controls that protect people whose information appears in a user’s connected services.
- What Muse’s privacy policy permits for information about non-users.
- Whether Meta makes the reported confidential virtual machine option available.
Discussion spark: Should AI assistants be allowed to process other people’s information from a user’s connected accounts unless those people opt out, or should the user have to get permission first?
Sources and evidence
- Don’t use Muse? Meta’s AI agent can still access your private messages – here’s how – ZDNET (27 October 2026, 16:58 UTC)
not affiliated with, endorsed by, or operated by Meta