Skip to content
Submit incident
AII-1000-54217
ReportedReported by fewer than three independent publishers, with no official finding.

Meta's Muse AI agent raises privacy and security concerns

Meta launched Muse, a personal AI agent that can browse the web, access user accounts, send emails, make purchases and perform other tasks on behalf of users. Privacy advocates raised concerns about Meta's history of data misuse, the agent's access to sensitive user data, and the lack of end-to-end encryption protections.

8 September 2026 ยท Meta ยท Privacy and surveillance
8 September 2026
AII-1000-542172 sources โ†“
Quick facts
Organization
Meta (subject)
Category
Privacy and surveillance
Country
United States
Source
engadget
Date
8 September 2026
Drafted by
Raidu Incident Detection

What happened

Meta launched Muse on September 8, 2026, a personal AI agent available on iOS, Android, web and WhatsApp that can browse the web, send emails, book travel, make purchases, and access connected third-party accounts including Google and Spotify.

The agent runs on Muse Spark 1.3 inside an isolated virtual machine, but Meta can still access the data. A promised Confidential VM that would cryptographically prevent Meta's access is "coming later in 2026," according to reporting.

Reported impact

Privacy advocates noted that Meta's history includes a $5 billion privacy fine, data monetization practices, and youth harm settlements, creating concerns about the company's handling of sensitive user data that the agent can access including email, calendar, bank accounts, and health data.

Users can opt out of allowing Meta to train its models on their interactions with Muse through data controls settings, but must actively change this setting from the default.

Index analysis

The agent is designed to require no technical experience and emphasizes accessibility, potentially reaching users who may not fully understand the security and privacy implications of granting the AI access to their financial accounts and personal data.

Governance controls

Meta's default permission setting is for Muse to "ask for some actions" but act autonomously on "low-risk" scenarios, though this relies on Meta's AI to distinguish what is actually low-risk. Users can change the permission setting to "always ask" but may experience increased friction in the agent's operation.

Classification

Organization
Meta (subject)
Category
Privacy and surveillance
Industry
Consumer services
Country
United States
Harm status
Potential harm
Intent
Unintentional
Timing
Unknown
Harm types
Economic or property
Harmed
Consumers
Consequences
None recorded
AI systems
Muse

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

engadget
How to get started with Meta's new AI agent, Muse - Engadget
12 September 2026
Memeburn
Meta Muse First Look: Inside the Web UI and Its Risks - Memeburn
13 September 2026