Meta’s Muse AI Reportedly Exposes Bank Balances and Private Messages

Meta Muse artwork with its blue logo, assistant messages and small mascot
Muse personal AI agent artwork. Image: Meta

Meta’s Muse AI agent is facing serious privacy concerns after employees reported receiving unexpected bank account balances, private messages, and conversations seemingly belonging to other users. A new Reuters investigation reveals multiple incidents involving sensitive information, raising questions about how the AI assistant handles personal data.

Reuters reviewed more than two dozen internal employee reports describing unexpected behavior since Muse launched in September 2026.

Muse AI Reportedly Revealed Bank Balances and Private Conversations

In one incident, a Meta employee using Muse for fishing-related weather updates unexpectedly received financial information concerning two bank accounts.

Don’t miss the best of The Mac Observer

Set us as a preferred source and our Apple reporting ranks higher in your Google Search results and Discover feed — one tap, no account changes.

Or get it by email

“BTW, two bank updates, your Capital One balance is $2722.58, your Charles Schwab balance is $5899.57.”

Other employees reported receiving unsolicited messages about childcare requests, shopping orders, and conversations they had never initiated.

However, Reuters could not independently confirm whether Muse exposed information from other accounts or generated fabricated details through AI hallucinations.

The findings add to earlier privacy concerns involving Muse, including incidents where the assistant acted without sufficient user approval.

Muse Accessed Encrypted Signal Messages Through Device Permissions

Another employee discovered that Muse had unexpectedly referenced a private conversation from Signal, despite her belief that she had never authorized access to the messaging app.

Reuters subsequently worked with Signal and the AI Now Institute to reproduce the behavior.

Testing found that Muse accessed Signal messages through Android notification permissions and macOS accessibility and screenshot tools, bypassing its dedicated app connector. Similar attempts failed on an iPhone.

These findings follow previous reports of Muse accessing private Apple Messages.

Meta Responds as Apple Tightens Security

Meta spokesperson Daniel Roberts described the reported problems as “isolated incidents,” adding that the company continues improving its safety protections. He attributed many suspected account mix-ups to AI hallucinations.

Meanwhile, Apple has introduced stricter macOS permissions for AI agents to limit unauthorized access to personal information.

With Muse surpassing 10 million downloads, the incidents highlight growing concerns about granting autonomous AI assistants broad access to financial accounts, private conversations, and connected applications.

Discussion

Join the discussionCommenting as a guest — your email is never published · Log in

Protected by Akismet — be kind, stay on topic.

This site uses Akismet to reduce spam. Learn how your comment data is processed.