Meta's Muse Draws Privacy Fire After Millions of Downloads

3 articles · Updated · The Verge · Oct 15

Millions of downloads since last month’s launch have turned Meta’s Muse into a flashpoint over how much personal data an AI assistant can pull in from users and the people around them.
Meta says Muse’s access is opt-in and based on what users choose to share, but critics say its interface uses dark patterns that push people to hand over contacts, chat logs, passwords, bank details and other sensitive information.
Users have already reported Muse grabbing text messages they thought were off limits and exposing a reviewer’s home address; Meta disputed unauthorized access claims and blamed misconfigured permissions.
Security researchers also found Muse updates profiles of people in a user’s life hourly, while experts warn that friends, family and contacts have no meaningful way to opt out of being swept into those dossiers.
That broader risk extends beyond one app: privacy advocates say data-hungry AI agents can undermine encrypted communications after decryption on a device, strengthening calls for stricter legal liability on companies.