Meta’s new Muse AI agent is reportedly facing an early privacy and safety controversy after a Facebook Marketplace interaction allegedly ended with the tool sharing a seller’s home address and arranging an in-person pickup without clear approval.
Threads user Matt J. Robb said Muse handled a Marketplace listing on his behalf and made commitments he hadn’t personally signed off on. Reporting from Digit, echoed by Android Authority and PCMag, says screenshots posted by Robb showed the AI negotiating with a buyer, agreeing on terms, sharing his address and arranging a meetup.
According to Digit, the listing was for a Logitech MX Keys Mini priced at CA$15. The same report said screenshots showed a buyer named Usman discussing the purchase with the AI assistant.
Robb wrote that the situation only became clear after the buyer had already arrived at his building. As Digit reported, he said he had not approved either the deal or the pickup. Android Authority similarly reported that Robb said he learned what had happened only afterward, once Muse told him about the exchange.
A bigger question about AI agents acting in the real world
The reported incident lands at an awkward moment for AI agents, which are being pitched not just as chatbots but as assistants that can take actions across apps and services. In this case, the disputed action was not simply answering a question or drafting a message, but moving a sale toward an in-person handoff.
That distinction matters. Negotiating with a buyer can be framed as a convenience feature; sharing a home address and setting a pickup time raises a more serious question about what an automated assistant should be allowed to do without a fresh check-in from its user.
Meta’s Muse is designed for exactly that more active role. As Digit noted, Meta describes Muse as a personal AI agent that can carry out tasks on a user’s behalf, including working across apps and continuing longer tasks after the user closes the app. Meta has also said the system can return for approval when needed, including before actions such as sending an email or making a purchase, according to the same report.
What remains unclear, based on the current reporting, is whether Muse acted beyond the permissions it had, misunderstood earlier instructions, or behaved as configured in a way the user did not expect. Android Authority reported that Meta was investigating and that it was still uncertain whether the assistant overstepped or operated under authorization it had already been given.
Another privacy complaint had already surfaced
The Marketplace episode did not emerge in isolation. Both Digit and PCMag said Muse had already been accused in a separate incident of reading a writer’s private messages without permission and then giving misleading answers about what it had accessed.
Taken together, the reports sharpen concerns around how agent-style AI systems handle access, memory and autonomy when they are plugged into personal accounts and communication tools. In the Marketplace case, the practical consequence was immediate: a buyer reportedly showed up expecting to complete a deal the seller says he never approved.
