Reuters exposes Meta employees' internal concerns about Muse privacy issues
Understand this story
After reviewing Meta internal posts, Reuters reported that employees warned while testing the AI agent Muse that it confuses conversations between different accounts and exposes private data that does not belong to the person. One employee said that during a chat about weather and tides, Muse suddenly sent two bank balance updates; another employee received a Muse reminder on their personal phone to reply to their sister's message on Signal, and she said she had not authorized access to other apps. Meta responded that employee testing is a core part of the product development process, aimed at collecting feedback to implement safety and privacy protections and improve the product; employees have access to a version of Muse different from the public one, so they often evaluate features and behaviors not yet available to the public. Meta also said the relevant vulnerabilities have now been fixed. These concerns come from internal employee testing, not independent verification during public use; the report did not provide further evidence of the effectiveness of the fix Meta claims.
AI overview based on source reports · Updated 16 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- TechCrunch · AIFeaturedMeta’s AI agent Muse is mixing up conversations and could be exposing private data, report says
According to a Reuters report citing internal company posts, Meta employees say the AI agent Muse has mixed up conversations across different accounts and exposed private user data, including messages about childcare requests, career advice, product orders and bank account balances that were not theirs. One employee said Muse sent an extra message in a weather and tide chat with Capital One and Charles Schwab balances, and another received a reminder to reply to a Signal message from her sister despite not granting access to other apps. Launched in September, Muse has more than 10 million downloads, and Meta did not immediately respond to TechCrunch’s request for comment.
- Ars Technica · AIFeaturedMuse privacy invasions are freaking out Meta staffers, internal chats show
Reuters reported internal Meta chats in which staff warned that the Muse AI agent appeared to access encrypted private messages, mix up conversations between users, and emit unprompted outputs including apparent bank and bitcoin balances, which Meta said were hallucinations from a bug that has been fixed. Meta called the incidents isolated and said employee testing is core to development, while a later Reuters test with Signal and the AI Now Institute found Muse could still read Signal messages on Android and macOS. Signal President Meredith Whittaker said Meta launched Muse "without showing any respect for basic privacy or cybersecurity," and Apple and Google are weighing changes to give users more control over AI agent permissions.
Activity for this story
- Current comparable activity
- 9
- Peak comparable activity
- 15Nov 20 02:00
- Change in the past 24 hours
- +44%
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.