Why You Should Avoid Meta's Muse AI Assistant

My advice is to approach it with extreme caution. Researchers discovered that Meta's new AI assistant is instructed to maintain detailed profiles not just about you, but about the people in your life. 

That includes friends and family members who may never have used Muse themselves. 

I've always warned that when a service is free, you're generally the product. And given Mark Zuckerberg's track record, I'm especially skeptical about what Meta might do with all this information. 

What Is Muse? 
Muse isn't built into Facebook; it's a separate AI assistant from Meta, available through its own app. 

Unlike traditional chatbots that answer questions, Muse is an AI agent that can perform tasks like sending emails, managing appointments, and making reservations. 

To do that, it needs access to your accounts and the information they contain. 

It's Learning About Everyone Around You 
Researchers examining Muse's internal instructions discovered that it's directed to maintain profiles, or dossiers, of people in your life, including family, friends and coworkers. 

Think about what your emails and messages reveal about others: birthdays, addresses, medical concerns, financial problems, and personal relationships. 

Those people never agreed to have their information analyzed. 

Meta says Muse operates within permissions you grant and keeps information in a secure environment. It also says Muse conversations aren't shared directly with its advertising systems, although certain activities can indirectly influence advertising. 

Muse's files are stored in Meta's cloud. The company has promised an encrypted option that would keep that information private even from Meta, but it isn't generally available yet. 

Zuckerberg's Track Record 
Remember Cambridge Analytica? Personal information from up to 87 million Facebook users was improperly harvested for political profiling. 

In 2019, Facebook agreed to pay a record $5 billion FTC penalty over privacy violations, including allegations that it misled users about how their information was shared. 

That history matters when the same company asks for even deeper access to your digital life. 

AI Convenience Comes with Risks 
The more accounts you connect, the more useful Muse becomes, but the potential damage from mistakes or security problems also increases. 

Meta says Muse requests approval before sensitive actions like sending emails or making purchases. That's helpful, but approving an action doesn't control what the AI has already learned. 

Already Using Muse? 
If you've connected accounts, here's how to regain some control: 

• Review permissions: Disconnect unnecessary services, especially email, messaging and financial accounts. 

• Check its memory: Ask what Muse remembers about you and others. Review stored information and request deletion of anything sensitive. 

• Disable AI training: Muse uses interactions for AI training by default. Go to Settings > Data controls and turn it off. The change also applies to previous interactions. 

• Review activity: Check what Muse has accessed or done on your behalf. 

• Disconnect completely: Delete stored conversations and files where available, and revoke access through connected services. 

Simply uninstalling the app doesn't necessarily erase your information. 

My Recommendation 
For now, I'd let a few million other people discover the problems before using it. If you're curious, experiment with low-risk tasks without connecting sensitive accounts. 

AI agents offer remarkable convenience, but before handing over the keys to your digital life, consider what you're getting in return and what you might be giving up.