Meta's Muse AI agent allegedly exposing private data and mixing up conversations
Tech Desk
Meta staff members have voiced worries that the company’s AI agent, Muse, is conflating conversations between various accounts and revealing private user information, a new report based on internal company posts has revealed.
Advertisement
Instances highlighted by employees include Muse displaying messages regarding childcare requests that were never made, unsolicited career guidance, unplaced product orders, and bank account balances belonging to other people.
One staff member reported that Muse unexpectedly inserted a message into a weather and tide chat, stating: “BTW, two bank updates, your Capital One balance is $2722.58, your Charles Schwab balance is $5899.57.”
In a separate incident, an employee testing the AI agent on her personal mobile device was prompted by Muse to reply to a message from her sister on the end-to-end encrypted messaging service, Signal.
“This seems like an extreme privacy violation. I definitely did not grant it access to other apps, or recall being asked about this,” the employee noted in an internal post.
Advertisement
Muse, which debuted in September, has become one of the fastest-growing applications in history, amassing over 10 million downloads. Meta failed to provide an immediate response to a request for comment regarding these claims.