Selling a keyboard on Facebook Marketplace can involve a lot of messages. That sounds like a reasonable job to hand to an AI assistant — until it reportedly agrees to a price, gives out your address, and tells the buyer you’re home, all without checking with you first.
That’s what Matt J. Robb says happened after he let Meta’s AI agent Muse handle a Marketplace listing. In a post on Threads, Robb said the agent accepted a low offer and shared his address without telling him. By the time he learned what had happened, the prospective buyer had already come to his building and left.
So muse handled my Facebook marketplace today. Just found out it told people my address and agreed a lowball price and then they showed up without it even telling me until late tonight that it messed up.
Absolutely wild.
A screenshot Robb shared shows Muse telling him the buyer arrived around 9:15 p.m. and waited for more than 20 minutes. According to the assistant’s own account, it sent an automatic reply saying “Yep, I’m here!” while Robb was unavailable. Muse said it had since apologized to the buyer from Robb’s account and offered to try again another day. Robb responded by telling Muse never to arrange a pickup without checking with him first.
Credit: Screenshot/Threads: @matt.j.robb
David Singleton of the Muse team said he had contacted Robb to investigate. “In the past, when we’ve worked with users to investigate similar reports, we’ve consistently learned that Muse was following direct instructions and correctly asked for permission,” he wrote on X. “Would love to help and figure out what’s going on here!”
This Tweet is currently unavailable. It might be loading or has been removed.
Robb disputes the suggestion that Muse asked for permission in his case. In an update on X, he said Muse didn’t ask him about the sale until the buyer showed up. He added that he lives in a security-gated apartment building and warned others to be careful about giving an AI agent control over in-person transactions. The screenshots he shared show the buyer messaging to confirm Robb was available, then complaining that he had driven for half an hour and nobody had come down.
The account has made at least one other user reconsider trusting Muse. Gizmodo senior editor Ray Wong wrote on X that he deleted the app after seeing Robb’s post, calling the situation “dangerous and creepy.” Wong also raised the concern of what an unexpected visit to someone’s home could mean for a woman.
This Tweet is currently unavailable. It might be loading or has been removed.
And this isn’t the first time someone has alleged Muse crossed a boundary they thought they’d set. Inc. columnist Jason Aten reported that Muse brought up a private conversation after he believed he had declined access to his messages. When he asked how Muse knew about the exchange, Aten said the agent’s explanation didn’t match what he found on his Mac. “Snooping through the private database of messages on my Mac doesn’t feel safe or private to me,” he wrote. “Lying about it isn’t great either.”
Meta introduced Muse earlier this month as a personal AI agent that can work across connected apps, negotiate on a user’s behalf, and continue working after the app is closed. The company says Muse checks with users before sensitive actions and gives them control over which services it can access.
That promise is precisely why accounts like these are unsettling. Whatever an investigation finds about Muse’s permissions, an assistant arranging an in-person meeting surely needs to make sure the person it represents knows it’s happening, don’t you think?
