Meta’s Muse AI agents are gaining traction in the U.S., with the app surpassing 5 million downloads last week. So it’s worth reviewing the information Meta is gathering to power its Muse recommendations, including the specific types of personal info that Muse can infer and track, based on people’s interactions with the app.
Meta’s Muse personal agents are designed to provide guidance on virtually every aspect of users’ lives, including giving suggestions on health, finances and relationships. In order to do that, Muse needs to take in more personal data than Meta’s other apps have ever needed. However, previous research has shown that Facebook likely knows more about users than their friends or family, so this intake of personal information seems like a potential concern.
In addition, Meta has a concerning track record of repeatedly misusing people’s data, which should make this a bigger consideration.
This week, a report from Wired once again underlined privacy concerns. Wired reported that security researchers extracted Muse operating system elements, including instructions on how the app should gather and use data.
The results are concerning.
As reported by Wired: “One of Muse’s instructions appears to be the ability for it to create ‘a page for every person in the user’s life.’ This hourly process involves compiling data on family, partners, friends, colleagues, ‘collaborators,’ and people you ‘follow,’ the instructions say. The general idea is that Muse can use its ‘memory’ (aka structured text files) to collect information about your relationships and the important people in your life.”
Based on this, Muse can provide recommendations on how to improve relationships. This was actually a key use case that Meta CEO Mark Zuckerberg highlighted in his initial discussion of the company’s AI superintelligence project.
Indeed, Zuckerberg has repeatedly said that AI can make people better parents, better friends and better partners. But in order to do that, Muse AI needs to map out each user’s personal connections and understand their relevance.
So what type of information does Muse track in order to build friend and family profiles?
Per Wired: “‘Where they live, what they do, the threads that recur (the apartment move, the shared savings goal),’ the instructions say. They add, too, that the model could record the ‘dates that matter,’ such as birthdays or anniversaries. A person’s history could include backstory like ‘the trip in March, the argument that got resolved, the milestone last week.’”
If the aim is to use Muse as a relationship guide, this type of data gathering makes sense. But even so, that’s a lot of very personal info that Meta is now keeping a log of for each Muse user.
This could be a problem
It’s worth reiterating that Meta makes 97% of its revenue from advertising. That means at least some portion of these insights will be used for ad targeting, which could lead to all sorts of questionable approaches based on what could amount to psychological vulnerabilities.
There are also concerns as to how that data could be used to angle political coverage or prey on users’ vulnerabilities to influence election outcomes.
Maybe Meta could use these insights to keep users hooked to its apps, or maybe Meta’s data could be leaked.
Clearly, past lessons of data misuse haven’t yet informed Muse take-up, since downloads of Muse are growing faster than ChatGPT at the same stage.
Meta has consistently been able to counter privacy concerns with utility, and it may do so yet again with Muse. But the issue is worth following.
