Meta's AI agent Muse mixes conversations and exposes personal data
Meta employees reported that the AI agent Muse sometimes mixed data from different accounts and displayed other users' personal information. This was revealed in an investigation by Reuters.

Reuters reviewed internal Meta messages in which employees described issues with Muse. According to them, users saw queries about children, career advice, and orders they had not placed, as well as other people's bank balances.
In one case, the agent appended a message containing account details from Capital One and Charles Schwab, belonging to another person, to a conversation about weather and tides. In another instance, Muse reminded a female employee on her personal phone to reply to her sister in Signal—a end-to-end encrypted app. She wrote to colleagues that she had not granted the agent access to other applications and did not recall being asked about it.
Muse was launched in September. With over 10 million downloads, it became one of the fastest-growing apps. Meta did not immediately respond to a request for comment.




