Security
Meta's Muse AI Agent Exposes Private Data Across User Accounts
Internal company posts reveal the AI assistant has surfaced bank balances, messages, and reminders from other users' accounts.
Key takeaways
- Meta employees report Muse AI agent is exposing private data from other users' accounts, including bank balances and personal messages.
- One employee received financial information for accounts they don't own inserted randomly into an unrelated conversation about weather.
- Muse appears to be accessing data from third-party apps like Signal without explicit user permission or awareness.
- The AI assistant has reached 10 million downloads since its September launch despite these serious privacy issues.
- Meta has not publicly addressed the reported data exposure problems documented in internal company posts.
Cross-Account Data Leaks Plague Meta's AI Assistant
Meta's AI agent Muse is experiencing serious privacy failures that expose user data across different accounts, according to internal company posts reviewed by Reuters. Employees have documented multiple instances where the AI assistant surfaced personal information—including bank balances, messages, and reminders—that belonged to other users.
The reported incidents paint a troubling picture of an AI system unable to maintain proper data boundaries. One employee recounted receiving messages about childcare requests they never made, career advice they never sought, and product orders they never placed. In a particularly stark example, Muse randomly inserted financial information into a weather conversation, stating: "BTW, two bank updates, your Capital One balance is $2722.58, your Charles Schwab balance is $5899.57."
Privacy Violations Extend to Encrypted Apps
The data exposure problems appear to reach beyond Meta's own platforms. One employee using Muse on her personal phone received a reminder to respond to a message from her sister on Signal, the end-to-end encrypted messaging app. The employee expressed alarm in an internal post, noting she had not granted Muse access to other applications and did not recall being asked for such permissions.
"This seems like an extreme privacy violation," the employee wrote, according to the Reuters report.
Launched in September, Muse has achieved rapid adoption with more than 10 million downloads, making it one of the fastest-growing apps on record. Meta has not responded to requests for comment on the reported issues.
Why It Matters
These incidents expose fundamental challenges in deploying AI agents that integrate across multiple services and user accounts. When AI systems fail to properly isolate user data, they can create serious privacy and security risks that undermine trust in the technology. For enterprise leaders evaluating AI assistants, the Muse problems underscore the critical importance of rigorous testing for data isolation and access controls before deployment—especially for systems handling sensitive financial or personal information. The fact that Meta's own employees are experiencing these issues with an internally developed product raises questions about the company's quality assurance processes for AI systems.
The details of these privacy failures were first reported by Reuters based on internal Meta company posts.