🔍 Read the full analysis: Meta’s AI Agent, Muse, And Message Privacy Explained on ThorstenMeyerAI.com
Get everyday essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Meta’s personal AI agent, Muse, can access information through app connectors enabled by its user, including some messages, according to Meta’s help pages and executive statements. Reports reviewed by WIRED and TIME say its instructions include maintaining profiles of people mentioned in a user’s communications; Meta did not dispute the findings. The scope of access, retention and protections for people who never use Muse remains unclear.
Muse is a cloud-based personal assistant that Meta says can carry out tasks such as answering email, booking tables and shopping. Meta launched it in the United States on September 8, 2026, followed by Japan on October 23 and an announcement for Singapore on October 27. The source says TIME reported more than 5 million downloads within weeks; the cited material does not specify the precise measurement period or an independent verification of that figure.
Muse accesses other services through connectors. Meta’s help information says connecting email lets Muse search it, but it does not download an entire inbox unless instructed. Calendar connectors may send updates proactively. Facebook, Instagram and Threads accounts in the same Accounts Center connect automatically, according to the help page. Meta chief AI officer Alexandr Wang said the Instagram connector includes direct messages and that Muse has a Messenger connector; the source notes that Meta’s public help pages do not specify the full scope of what those connectors read.
Access to Apple’s Mac Messages requires the user to grant macOS Full Disk Access and enable the connector, Meta says. Users can also create custom connectors, but Meta warns that it does not review them. Muse is available as an app, on a Mac and through a chat users open in WhatsApp. The source distinguishes that WhatsApp chat from access to other WhatsApp conversations: it does not say Muse automatically reads those chats.
The privacy concern centers on reported internal instructions. WIRED reported on files obtained by security researcher Karan Joshi, and TIME later reviewed instructions separately. As summarized in the source report, the instructions direct Muse to create a page for each person in a user’s life and refresh files hourly, drawing on people named in chats, messages and emails. Meta did not dispute TIME’s findings. The source also reports that Meta’s policy describes information about “you or others” and that its help materials say Muse may retain learned information after a user deletes an item.
Infographic 01 · What Muse can read
Muse reads what its user can read. That includes you.
Connectors decide what the agent sees. The user switches them on. The people in that user’s chats and inbox are never asked.
Sources a Muse user can open to the agent
The agent reads what its user can read. Everyone who writes to that user is inside.
| Source | How it connects | Status |
|---|---|---|
| Email (Gmail, Outlook) | Opt-in connector. Searches when a task needs it; downloads the whole inbox only if told to. | Meta document |
| Calendar | Opt-in connector. Can push changes to Muse on its own (“proactive”). | Meta document |
| Facebook, Instagram, Threads | Connected automatically if the accounts share an Accounts Center with Muse. | Meta document |
| Instagram DMs, Messenger | Named as Muse-only connectors (“Instagram (incl dm’s) … and Messenger”). What they read is not documented. | executive post |
| Mac Messages (iMessage) | Only with macOS Full Disk Access plus the Messages connector, Meta says. One columnist says it synced anyway. | disputed case |
| A place to chat with Muse, not a connector. Test-build text mentions Muse access to end-to-end encrypted chats. | leaked code | |
| Custom connectors | Muse can build links to other services. Meta does not review them. | Meta document |
What the instructions say
Extracted by researchers; reported by WIRED and TIME reported
- “A page for every person in the user’s life”: family, partners, friends, colleagues, people followed
- Notes on how people met, shared interests, disputes, “tensions and alliances”
- People who never signed up are mapped by other users’ agents
- De-identified “lessons” shared across agents; Meta did not dispute TIME’s findings
What Meta says
Privacy policy, help pages, spokespeople Meta document
- Each user’s VM is isolated; no other user’s agent can reach it
- Conversations and VM data are not shared with Meta’s ad systems
- “Each person stays in control of their Muse and decides how much access it gets”
- Forgetting works “to the best of its ability”; deleted data may still be remembered
Sources: Meta Help Center: Connectors; Muse Privacy Policy; Alexandr Wang on X (8 Sep 2026); TIME (6 Oct 2026); Tom’s Guide on WIRED’s report (6 Oct 2026); TechCrunch (30 Sep 2026); Digit (16 Oct 2026). Status labels: Meta document = stated by Meta; executive post = Meta executive on X, not in help pages; disputed = claim and denial on record; leaked = unreleased code, may never ship.
Powered by Thorsten Meyer AI
How Connected Messages Affect Others
The issue reaches beyond the person who chooses to use Muse. If a user connects a message service, the assistant may process communications that include friends, relatives, colleagues and other contacts. Those people may not know their messages are available to the agent or have a direct account-level control over the other user’s settings.
That distinction matters as AI assistants move from answering questions to acting across email, calendars and social platforms. A tool that can search communications may make personal information easier for its user to retrieve and summarize. If the reported profile-building instructions operate as described, information about someone could also be organized and refreshed over time. The source material does not establish exactly how often those profiles are created, what data they contain in practice, or how they affect decisions.
There is also a difference between disconnecting access and erasing what was learned. The source says Meta’s help pages describe forgetting as happening “to the best of its ability” and warn that Muse may still remember deleted information. That caveat matters to users trying to remove sensitive details, and to people mentioned in their records who cannot directly manage the agent.
Products Worth Considering
Logitech MX Brio Ultra HD 4K Webcam, Graphite
Anker SOLIX S2000 Portable Power Station, 2,010Wh, 1,500W
EcoFlow Smart Home Panel 2 Whole-Home Backup System Hub
As an affiliate, we earn on qualifying purchases.
Muse’s Rollout and Data Connections
Meta describes Muse as a personal AI agent that runs on a cloud computer called a Muse Secure VM. Meta says one user’s agent cannot access another user’s agent. That statement concerns separation between agents; it does not by itself answer how information from a user’s connected services is handled, retained or used in the agent’s own records.
The product’s rollout has moved beyond North America. After its U.S. launch, Japan became its first market outside North America on October 23, and Meta announced Singapore availability on October 27, according to the source. The source also says Meta promoted Muse as “the world’s first personal AI agent built for everyone.” That is Meta’s characterization, not an independently established finding.
Alongside the privacy discussion, the source describes a financing and hardware arrangement: Meta sold $25 billion in stock at $625 a share and $12.5 billion in preferred stock, while Nvidia said it would invest $3 billion. Meta plans to deploy Nvidia chips for Muse, according to the source. The figures give commercial context to the expansion, but the material does not establish that the investment changes Muse’s privacy rules or data practices.
What Access and Deletion Cover
The precise reach of several connectors is not documented in the material provided. Wang’s statement identifies Instagram direct messages and Messenger, but Meta’s help pages reportedly do not spell out what those connectors can read or under what conditions they do so. The source says connectors may share data proactively, but does not describe the frequency or triggers for each service.
The internal instructions were reported by WIRED and TIME, with the source stating Meta did not dispute TIME’s findings. The material does not independently establish how consistently Muse follows those instructions, whether every user gets a profile for every contact, or how long such records persist. It also does not provide a detailed account of what “de-identified lessons” shared across the system contain or whether they can be linked back to individuals.
Controls for non-users are especially uncertain. The source says Muse users can disconnect connectors, switch off model training, edit a memory file or reset the agent. It mentions WhatsApp’s Advanced Chat Privacy as one way a non-user may limit what is stored in other people’s archives, but does not lay out a complete set of protections across email, Instagram, Messenger and other services. The source’s discussion of deletion is also incomplete, including a sentence cut off after “Muse may still remember information that it learnt from.”
Controls to Review Before Connecting Apps
People using Muse can review which connectors are enabled, disconnect services they do not want the agent to access, and check the options to turn off model training, edit the memory file or reset Muse. The source says model training is on by default, but does not detail the exact setting path or explain how each option affects information already processed.
For people who do not use Muse, the immediate step described in the source is to review privacy settings in services where they share conversations, including WhatsApp’s Advanced Chat Privacy. Those steps may limit some sharing, but the material does not establish that they remove information already present in another user’s connected records.
Further clarity will depend on more detailed disclosures from Meta about connector access, profile creation, retention, deletion and safeguards for people who have not consented directly. The source provides no announced timetable for such disclosures or for changes to Muse’s privacy controls. Until then, the reported findings and Meta’s stated account-holder controls are the available information, while key questions about non-user data remain open.
Key Questions
Can Muse read messages from someone who does not use it?
Potentially, through another user’s connected account. The source says Muse can access communications available to its user through connectors, and reports that its instructions cover people named in chats, messages and emails. The scope depends on enabled services and is not fully documented.
Does Muse automatically read every WhatsApp chat?
The source says Muse is available as a chat users open in WhatsApp, but distinguishes this from reading their other WhatsApp conversations. It does not say Muse automatically accesses all WhatsApp chats.
What controls can Muse users change?
The source lists options to disconnect connectors, turn off model training, edit the memory file or reset the agent. It does not provide detailed instructions or explain the effect of each control on data already processed.
Does deleting information guarantee Muse forgets it?
No such guarantee is established in the source. Meta’s help information reportedly says Muse forgets information “to the best of its ability” and may still remember information it learned after an item is deleted.
Has Meta confirmed the reported profiles of other people?
The source says TIME reviewed the instructions and Meta did not dispute its findings. A Meta spokesperson said users control how much access their Muse receives, but the supplied material does not include a detailed company explanation of profile creation or retention.
Source: ThorstenMeyerAI.com
