🔍 Read the full analysis: What Does Meta’s AI Agent See—and How Can You Limit It? on ThorstenMeyerAI.com
Get the latest gadgets 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 Muse personal AI agent can access information through connected services, including some Meta accounts that share an Accounts Center. Reports say its internal instructions direct it to keep profiles of people mentioned in a user’s communications; Meta did not dispute those findings. Users can manage Muse’s access, but people whose information appears in someone else’s messages have fewer direct controls.
Meta announced Muse’s expansion to Singapore on October 27, while reporting has raised questions about how the personal AI agent handles information about people who never signed up. Meta’s documentation describes access through connected services, and two investigations reported that Muse is instructed to keep profiles of people mentioned in a user’s communications; Meta did not dispute the reported findings.
Muse uses connectors to access services on a user’s behalf. Meta says email access lets Muse search messages but does not download an entire inbox unless asked; calendar connectors can send updates proactively. Meta accounts in the same Accounts Center as Muse may connect automatically, according to Meta’s help information. Meta chief AI officer Alexandr Wang has also said the Instagram connector includes direct messages and that Messenger is supported, though the precise scope is not fully described in the help pages.
Other access depends on permission. Meta says its Mac Messages connector requires the user to grant macOS Full Disk Access and enable the connector. Users can also create custom connectors, but Meta says it does not review them. Muse is available through its app and on Mac; Meta says it can also be opened as a chat in WhatsApp. That is distinct from reading a user’s other WhatsApp conversations.
WIRED and TIME reported that Muse’s internal instructions tell it to create and update a page about each person in the user’s life, including people identified in chats, messages and emails. TIME said the pages may include shared history and relationship details and are refreshed hourly. Meta did not dispute those findings. Its policy says Muse may store “information about you or others,” while the company says users decide what access their agents receive.
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
When Someone Else’s Agent Has Your Messages
The privacy issue extends beyond the person who installs Muse. If a user connects an account or gives an agent access to communications, details about family members, colleagues and friends may be processed as part of that user’s tasks. Those people may not know an agent can search or summarize the material, and they generally cannot control the other person’s Muse settings.
That creates a difference between account control and information control: Meta says a Muse user controls the agent, but people described in the user’s correspondence may have little say over how their information is handled. The effect depends on what the user connects, what the agent is asked to do, and what information it retains.
The commercial stakes are also growing. Meta announced Muse’s Singapore launch as it expands the product internationally. The source reports a $25 billion stock sale at $625 a share, $12.5 billion in preferred stock, and a $3 billion Nvidia investment. Nvidia’s chips are to be used by Meta for Muse, according to the source. The investment does not establish what data Muse collects, but it underscores the scale of the companies’ partnership as Meta develops the agent.
Popular right now
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
KitchenAid Artisan Series 5-Quart Tilt-Head Stand Mixer, Dried Rose (KSM150PS)
As an affiliate, we earn on qualifying purchases.
Muse’s Rollout and Data Connections
Meta launched Muse in the United States on September 8, 2026, and Japan followed on October 23, according to the source material. Singapore, announced October 27, is its first market in Southeast Asia. Meta describes Muse as a personal AI agent that can perform tasks such as answering email, booking tables and shopping. Each user receives a “Muse Secure VM,” which Meta says is a private virtual computer that other users’ agents cannot access.
The reporting about profiles followed examinations of Muse’s instructions. WIRED reported on October 3 that security researcher Karan Joshi obtained files by asking Muse about them. TIME reviewed the instructions separately three days later and reported hourly updates to files about users and people mentioned in their communications. The accounts of those investigations are the basis for claims about the profile system; they are not the same as a public, detailed account from Meta of every data flow.
Meta’s own help materials describe practical settings: users can disconnect connectors, switch off model training, edit the memory file or reset Muse. The source says training is on by default. Meta’s guidance also cautions that asking Muse to forget information may not remove what it learned from the original content.
“information about you or others”
— Meta privacy policy
Limits of Muse’s Memory Controls
The full scope of data access remains unclear. Meta’s public help pages do not spell out precisely what Instagram DMs or Messenger data their connectors can read, how long information obtained through every connector is retained, or how often it is accessed. The source says connectors can share information proactively, but the timing and circumstances may vary by service and task.
Meta’s help material says Muse may still remember information learned from content a user deletes and describes forgetting as something it does “to the best of its ability.” That means a request to forget should not be treated as proof that the original messages or derived details have been deleted. The source also reports that TIME found instructions discouraging Muse from pointing out that original messages may remain visible; Meta has not provided a detailed public account addressing that specific point.
It is also unclear how profiles about non-users are stored, how long they persist, whether a person can request access to or removal of a profile created by someone else’s agent, and how reported de-identified lessons are used. Meta’s statement that the user controls their Muse does not resolve those questions.
Settings Users Can Review Now
Muse users can start by checking which connectors are enabled, disconnecting services they do not need, and reviewing whether accounts share an Accounts Center. They can also switch off model training, edit Muse’s memory file or reset the agent. For email, Meta says users can ask Muse to search without authorizing a full inbox download unless they specifically request one. Users should treat custom connectors cautiously because Meta says it does not review them.
People who do not use Muse have fewer direct options over information held in someone else’s account. The source points to WhatsApp’s Advanced Chat Privacy as one way to limit what can be taken out of a chat, though that does not control other services or erase material already available to another person. Users can also ask contacts not to share sensitive conversations with AI tools, but that depends on those contacts’ choices.
The next key development is whether Meta publishes clearer connector-by-connector explanations and specific retention and deletion rules, particularly for information about non-users. Meta’s international rollout is continuing, but the source does not give a date for additional markets or a timetable for further privacy disclosures.
Key Questions
Can Muse read every message in a user’s accounts?
Not necessarily. Access depends on enabled connectors and the service. Meta says email search does not download an entire inbox unless the user asks, while some Meta accounts may connect automatically through a shared Accounts Center. The exact scope of certain messaging connectors is not fully documented in the help pages.
Can Muse create information about people who do not use it?
WIRED and TIME reported that Muse’s instructions tell it to maintain pages about people mentioned in a user’s communications. Meta did not dispute those findings. The person described in a page may not have a Muse account or direct control over the other user’s settings.
Does asking Muse to forget something delete it?
Not necessarily. Meta’s help information says Muse may retain information learned from deleted content and describes forgetting as best-effort. Users should not assume that removing a memory erases the original messages or every detail derived from them.
How can a Muse user limit what it can access?
Users can review and disconnect connectors, turn off model training, edit the memory file or reset Muse. They can also avoid connecting accounts or services that are not needed and review any custom connectors, which Meta says it does not review.
What can non-users do about information in someone else’s chats?
They have fewer direct controls over another person’s Muse. The source cites WhatsApp’s Advanced Chat Privacy as a way to limit what can be taken out of a chat, but it does not cover all services or remove information already available to another participant.
Source: ThorstenMeyerAI.com
