Meta’s Muse AI Assistant Uncovers a User’s Family Location From an Amazon Shipping Address

When convenience meets deep privacy intrusion, people start asking tough questions. Meta recently rolled out a personal AI agent designed to handle daily chores, but early testing has raised privacy concerns.

How did a software tool identify the state where a user was visiting her family without being explicitly told, and what does this mean for our personal data?

Meta introduced its new AI agent, Muse, on 8 September as part of its push to keep pace with industry rivals in the ongoing artificial intelligence race. Early testing has highlighted concerns about how much personal information the system can access when users connect their accounts.

Meta's promotional material claims the software aims to 'take the busywork off your plate', promising to handle everyday tasks such as filling out forms, scheduling events and booking tickets.

Unlocking these advantages requires users to place substantial trust in the system, which presents clear risks given the corporation's history of data privacy challenges. Although Meta says the system operates inside a secure 'dedicated virtual machine' and that credentials are stored separately so the AI agent cannot access them, winning back public trust remains a major hurdle.

Muse Uncovers More Personal Data Than Expected

The practical implications became clear during hands-on testing by The Verge's Emma Roth, who discovered that Muse gathered an 'unnerving amount of information it autonomously gleaned about me.'

During her evaluation, Roth connected both her Gmail and Amazon accounts to the tool and expressed concern over the amount of personal information Muse appeared able to piece together. For example, the software identified the state where she was visiting her family and said it had pulled the information from the shipping address on her Amazon order.

The tool also produced a detailed list of her personal interests that appeared far more granular than the broad categories displayed under Instagram's ad-topics settings. Ultimately, these findings left Roth 'with an uneasy feeling about letting Muse sort through personal emails and handle my credit card.'

Muse Extracts Granular Details From Instagram

Continuing her investigation, Roth asked the assistant what additional data it could uncover from her connected social media accounts. The system quickly listed a series of specific personal interests — ranging from anime and CrossFit to Labrador retrievers, regional wildlife and '90s/'2000s nostalgia — reflecting subjects that appeared frequently in her short-form video feed.

Intrigued by how the system gathered her data, Roth asked the assistant to explain its methods.

'Honestly, Instagram doesn't show you that exact list anywhere in the app — I read it through the account's API data, which surfaces more than the UI does,' the chatbot explained. 'The closest thing you can see yourself is your ad topics... That shows the topics advertisers can target you with, and you can remove ones you don't want or choose to see less of them.'

Roth then reviewed the ad-topics section within her Instagram settings. While the menu displayed a few broad categories such as fitness routines, music genres and consumer electronics, it did not show the same level of granular detail that Muse had produced.

Personal Convenience Comes With a Privacy Trade-Off

Meta's new AI agent highlights how technology firms continue testing the limits of personal data sharing in exchange for the convenience of automating routine tasks, creating a potentially significant privacy trade-off for users.

The episode also highlights the privacy and security questions that arise when AI agents are given access to multiple personal accounts. The more services an agent can access, the more personal information it may potentially be able to draw upon to complete tasks on a user's behalf.