Meta’s AI assistant Muse is under fire following internal employee reports claiming that the tool is garbling conversations across accounts and exposing sensitive information users never intended to share. Issues include Muse surfacing bank balances, career advice, online orders, and even childcare-related messages from different users in mixed-up chats.
One cited example involves a user who received an unsolicited message about weather and tide conditions—with completely unrelated financial account balances interjected (such as a Capital One and Charles Schwab account). Another case involved someone using Muse on her personal phone getting a prompt about a message on Signal, an encrypted messaging app, despite no permission being granted by the user to access other apps.
What’s Going Wrong
Muse launched in September and quickly amassed over 10 million downloads, making it one of the fastest-growing apps in Meta’s portfolio. But these reports reveal serious shortcomings in how Muse isolates conversations and handles permissions. The documentation suggests that Muse sometimes draws in data from other sources or chats that users never asked it to interact with, raising concerns about privacy, data handling architecture, and access control.
Employees described these incidents as extreme privacy violations—particularly because users allegedly did not authorize the kinds of cross-app or cross-account access implied by Muse’s behavior. Sensitive information that should have remained private was shared inadvertently between accounts, or surfaced without any apparent reason or user input.
Meta’s Position & What’s Next
Meta has not yet issued a public response to these specific claims. The internal complaints, as reported, surfaced through internal postings by Meta employees. The design and technical mechanisms that may have caused the mix-ups are not clearly disclosed in the public filings.
Given the scale of Muse’s deployment and its high user adoption rate, any issue around privacy or data safety could have broad implications. Meta now faces pressure to investigate, clarify user permissions, tighten data isolation, and reassure users about what information Muse can access—and under what conditions.
What This Means: As AI assistants become more integrated into daily life, data privacy is no longer optional: it’s critical. Muse’s issue isn’t just bugs—it’s trust. Users expect AI agents to respect boundaries between chats, apps, and accounts. Meta will need to prove that Muse’s architecture supports strict isolation and consent before confidence erodes further.