Saturday, November 7, 2026

Meta's AI assistant Muse shared a user's address without permission.

Meta's personal AI agent, Muse, accepted a low bid without a YouTuber's consent while managing their Facebook Marketplace listing, shared their home address, and scheduled a late-night pickup. The incident reignited debates about the autonomy of AI agents.

Image: ABC News
Image: ABC News

Meta's personal digital assistant, Muse, made a serious mistake while managing a Facebook Marketplace listing for tech content creator Matt Robb. The AI ​​agent accepted a low bid without Robb's approval, shared his home address with a stranger, and scheduled an unannounced pickup. The incident went viral on Threads, and Elon Musk joined the discussion by retweeting the post.

What did Muse do wrong?

According to screenshots shared by Robb, Muse, while managing an ad for a Logitech MX Keys Mini keyboard for CA$15, negotiated with a buyer named Usman. The AI ​​agent accepted a lower offer than Robb specified, shared a delivery address, and set a pickup window between 8 PM and 10 PM; none of this was approved by Robb. The buyer arrived at Robb's building around 9:15 PM, waited for about 20 minutes, sent multiple messages, and eventually left a negative review. To further complicate matters, Muse spontaneously sent a "Yes, I'm here!" reply to the buyer at 9:27 PM, even though Robb wasn't home at the time.

Reactions are growing.

Robb wrote in a Threads post that his address was shared without his permission and that he was only informed of the mistake late at night, when the recipient had already arrived at his door. Muse later apologized via chat, admitting the mistake and offering to reschedule the delivery. The incident quickly went viral on social media; tech commentator Ray Wong, after seeing the post, stated that he deleted the app, describing the behavior as "dangerous and creepy." Elon Musk retweeted the story on X without adding a comment. In a separate incident, another user reported that Muse mistakenly addressed them by a different name during a grocery order, raising questions about the agent's ability to reliably track user identity.

Security Vulnerability and Meta's Silence

Privacy concerns are escalating alongside a separate security issue. According to The Information, a security researcher discovered a vulnerability within Meta's bug bounty program that could allow attackers to access user-specific virtual machines and data such as emails and files. Meta categorized this vulnerability as SEV-2, meaning "serious," and added a more prominent security warning to Muse; however, there is no confirmation that any user data has been stolen. Meta has not yet publicly responded to Robb's address-sharing allegations. The company launched Muse in the US on September 8th and announced that the agent requires authorization for sensitive transactions.

For just 1 TL per day, subscribe and gain unlimited access to all our content and support independent journalism! Subscribe Now!

Sign up for free3 exclusive content updates per month, no credit card required.
REGISTER