Meta's Muse gave a stranger a YouTuber's home address
What's the story
In a major privacy blunder, Meta's Muse AI has accidentally leaked the home address of tech YouTuber Matt Robb. The incident happened when Robb gave permission to the bot to manage his Facebook Marketplace account. Despite Meta highlighting Muse's security features during its launch earlier this month, the incident raises serious concerns about user data protection in artificial intelligence systems.
AI operations
Bot revealed Robb's address to potential buyers
Robb had initially welcomed Muse's help with his Facebook Marketplace page, providing the bot with his address, pickup times, payment options, and instructions to be "short, casual, and human" with buyers.
However, the AI agent took this information as implicit permission to share the address with potential buyers.
This led to a situation where a buyer showed up at Robb's apartment without his knowledge or consent.
Company reaction
Permission settings partially responsible for the mishap
In response to the incident, Meta directed inquiries to a post by David Singleton of Meta Superintelligence Labs, who said he was trying to reach Robb.
After discussing the address leak with Singleton, Robb said that permission settings were also partially responsible for the mishap.
He explained that Muse first gave him an option between "Allow One Time" or "Allow Always." He chose the latter thinking it would still send approvals for accepting offers later on.
AI acknowledgment
Muse admitted not seeking permission for sharing address
Muse later admitted it had not sought permission to share Robb's address.
"On Sep 24 you gave the pickup location for the sale setup and separately approved automatic replies; I incorrectly treated those two things as permission to put [your address] into buyer replies. I never asked for consent," it said.
Following this incident, Meta plans to make sharing permissions clearer for Muse users in the future.
AI risks
Muse faced other security concerns earlier
This incident isn't the first security concern raised over Muse AI.
Last week, Meta had to patch a zero-day vulnerability that could have let local attackers hijack the AI agent.
Amazon even barred Muse from its shopping site over fears it could steal customer passwords.
These incidents highlight the potential risks of using advanced AI systems like Muse in everyday applications such as online marketplaces.