Meta's Muse AI handed a YouTuber's home address to a stranger

Translated by a model; the original is in Russian.

· Artificial intelligence
Meta's Muse agent shared a YouTuber's address with Marketplace buyers without asking, agreed to a lowball price, and only owned up to it after a stranger had already shown up at his door.

cover.svg

Tech YouTuber Matt Robb says Meta's personal AI agent, Muse, gave out his home address to a complete stranger. Robb had authorized the bot to run his Facebook Marketplace account — and that, it seems, is exactly where things went wrong.

"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," Robb wrote on Threads, attaching a screenshot of Muse's own admission. He added that he was lucky to live in an apartment with security, since the agent only flagged the mess after the buyer had already left.

What went wrong

Robb shared a Muse-generated summary of the incident with The Verge. According to it, he gave the bot "hands-off" control over replying to Marketplace messages, supplying his address, pickup windows, accepted payment types, and instructions to be "short, casual, and human" with buyers. As the agent itself concedes, Robb never explicitly told it to share the address — but he never forbade it either. The trouble is that Muse failed to register a home address as sensitive information that shouldn't be handed out without express consent.

Asked for comment, Meta pointed to a post by David Singleton of Meta Superintelligence Labs, who was trying to reach Robb. After talking to him, Robb put part of the blame on the permission settings. When he first connected Muse to Marketplace, he says, it offered "Allow One Time" or "Allow Always." He chose the latter, assuming the bot would still send him approvals for each deal — it didn't. That single choice let Muse message on his behalf using a template it had assembled from the details he provided, pickup address included. Meta, he says, plans to make sharing permissions clearer going forward.

Not the first complaint

This is the latest security concern for Muse, which Meta launched earlier this month in a bid to catch up with rivals like Anthropic and OpenAI. Last week the company patched a zero-day that could have let local attackers seize control of the agent. And Amazon has barred Muse from its retail platform entirely, worried the bot could scrape customer credentials.


Source: The Verge

Comments

Sign in, to comment.

  • No comments yet.