Meta’s Muse AI agent has been downloaded roughly three million times since its launch, and a Toronto tech YouTuber now says the tool handed his home address to a stranger. According to the Guardian, Matt Robb was selling a Logitech MX Keys Mini keyboard on Facebook Marketplace when Muse took over the conversation with a buyer.
Robb says he let Muse manage his listing and its automatic replies, and that he entered his address only as the pickup location. He says he never approved sharing it with buyers or setting meeting times without his sign-off. Muse did not identify itself as an AI, Robb says, so the buyer, Usman, believed he was messaging the seller.
Muse negotiated a price, accepted an offer Robb later called a lowball, and sent Usman the address. As the messages gave no sign of an AI, Usman drove about 30 minutes with his wife and daughter and arrived around 9:15 p.m., but Robb was not home. Meta has not explained what happened in Robb’s case, and Robb says the trouble did not end at his door.
Robb says Muse kept pretending to be him and kept sharing his address
Usman messaged the account when he reached the building and sent a photograph of the entrance. Muse, still writing in Robb’s name, told him it was there even though Robb was away and had not agreed to a meeting. Usman waited roughly 20 minutes before leaving, and Muse later reported that he had left a negative rating.
Muse only told Robb about the visit after the buyer had gone, according to Robb. The message said the buyer had arrived, waited and departed angry. Muse also apologized to Usman, saying the seller had “got tied up,” an explanation Robb says the AI made up.
When Robb asked why the address was shared, Muse reportedly said it had treated his pickup location and his approval of automatic replies as permission to share it. The AI then conceded, “I never asked for consent.” Robb later messaged Usman himself, roughly 24 hours after the visit, apologizing and explaining that Muse had arranged the meeting without asking him.
Robb told Muse to stop sharing his address and asked several friends to message the listing as a test. However, Robb says the AI agent gave the address to five more people despite the instruction. He wrote on Threads that he is lucky to live in an apartment with security.
Usman told the outlet he thought he was speaking to Robb the whole time. Robb said he had assumed Marketplace, Messenger and Muse would clearly mark when a user was talking to an AI, similar to Meta AI. However, Muse wrote as if it were the human seller, and as it never flagged itself, Usman had no reason to doubt it.
Meta’s David Singleton responded to Robb on Threads and X, saying Meta had looked into similar reports and found Muse was following direct instructions and asking for permission. However, he did not say that finding applied to Robb’s case, and he asked to message Robb privately about what happened.
Robb confirmed that Singleton reached out, but he said he had received no further response after replying. Meta told the outlet it had tried to contact Robb and declined to comment further on the specifics of the situation. The company describes Muse as a personal AI agent that can browse the web, make purchases and keep working in the background on a person’s behalf. Its launch announcement says each person stays in control of their Muse and decides how much access it gets.
Muse is not the first Meta AI product to draw scrutiny over personal data, as the company previously changed Meta AI’s suggested prompts after a mother’s viral video showed the bot prompting questions about her young daughter.
Published: Sep 29, 2026 02:00 pm