Signal

Meta’s AI agent Muse is mixing up conversations and exposing private data, report says

First reported by TechCrunch ·

The signal ●●○○ Compiled by AI from TechCrunch, Ars Technica, 404 Media, The Aisle, 9to5Mac and 5 more
Why you might care

Personal data accessed by Meta's AI agents is now visible across unrelated conversations, indicating a potential breach of user privacy and account security.

What happened

Meta employees have voiced concerns regarding the company's AI agent, Muse, which has reportedly been mixing up conversations and exposing private user data. Internal company posts reveal instances where Muse presented information unrelated to the user's queries or context, such as childcare requests, career advice, product orders, and financial balances from other accounts. One employee reported Muse randomly inserting bank balance details into a chat. Another employee expressed alarm after receiving a message from Muse prompting a response to a Signal message, indicating potential unauthorized access to other applications. Launched in September, Muse has experienced rapid growth, surpassing 10 million downloads.

What it means

Meta's AI agent Muse is exhibiting significant privacy flaws, with employees reporting incidents of crossed conversations and exposure of sensitive personal information like financial balances and private messages. This suggests a fundamental issue in how Muse is isolating and managing data across user sessions and potentially across different user accounts. The widespread nature of these reports, stemming from internal employee feedback, points to a systemic problem rather than isolated glitches.

The rapid adoption of Muse, exceeding 10 million downloads since its September launch, amplifies the risk associated with these data mishandling issues. As Meta integrates AI agents more deeply into its ecosystem, users may face a heightened risk of their private communications and personal data being inadvertently disclosed. This situation could lead to increased scrutiny from privacy regulators and a potential erosion of user trust in Meta's AI technologies.

AI-written summary. May contain errors.

AI Meta