Meta’s Muse AI Raises New Privacy Concerns Over User Data – ALM Corp

CALL US AT:
Get in Touch With Us
Home » News » Meta’s Muse AI Raises New Privacy Concerns Over User Data
Meta’s Muse AI agents are picking up users fast, and a new report is shining a light on the information the app collects to operate. Wired revealed this week that security researchers extracted portions of Muse’s operating instructions, which describe a procedure for constructing comprehensive pages about the people in a user’s life, according to the report, Social Media Today, by Andrew Hutchinson published October 4, 2026.
The findings aren’t a confirmed data breach.; They’re a look at how the app is told to behave. Still, for an assistant designed to advise on health, money and relationships, the details are worth reading closely.
Muse has experienced significant success since its launch. According to Social Media Today’s article, published on October 1, which cited Sensor Tower’s research data, the application scored over 5 million downloads. Moreover, the website states that only 23 applications managed to report more than 5 million downloads within their first 22 days. As per Yahoo Finance’s report, Muse is currently gaining more and more popularity, outperforming ChatGPT and xAI’s Grok in terms of engagement during its first few weeks on the market.
Furthermore, it is important to note that the application’s level of engagement seems to be consistently high. Indeed, according to The Information’s report, Muse currently boasts over 3 million weekly users, confirming that many people who downloaded the application continue to use it. In his recent article, Hutchinson notes that Muse’s growth rate appears to be higher than that of ChatGPT at a comparable period after its launch.
The publication by Social Media Today presents a list of updates introduced to the application since its launch on September 9. On September 20, Muse 1.0 for Mac was launched, while advanced features were unlocked for all users on September 28. As Hutchinson’s article explains, Muse represents the cornerstone of Meta’s strategy to enter the consumer AI market. Personal agents that will function as assistants in all areas of human life, from health and finance to relationships, will only be possible due to the information which will need to be entered into individual accounts. Therefore, as stated by Hutchinson, Muse will require a level of personal information that has never been collected by any other Meta platform before.
In the article, Wired magazine notes that security analysts examined parts of Muse’s operating system. In them, Hutchinson was especially interested in finding instructions for the application on how to collect and use data. The results of the analysis Hutchinson finds quite concerning.
One of the instructions seems to allow Muse to create for the user a page with information about every person in his life. According to Wired, the process is designed to be launched every hour. All the data about the user’s family, lovers, friends, colleagues, collaborators, and people they follow are collected into a single page.
The system, according to Wired, uses for this purpose Muse’s memory, which is stored in the form of structured text. It can be concluded from the instructions that Muse is capable of compiling this type of information about each user to analyze his social sphere and the people who are important to him.
It should be noted that this information comes from instructions found in the application’s operating system. Therefore, it refers only to the capabilities that Muse has, according to its creators. Thus, it cannot be argued based on this whether the application truly does all this frequently, whether it stores all this information on Meta’s servers, and for how long this information is stored there.
According to Wired, the instructions give further details on aspects like the residence and occupation of the people. Moreover, the instructions also note repetitive themes in the person’s life. For instance, the model can predict an apartment move or a shared savings plan in details.
The model can also observe and record important days depending on the person’s request. This could be birthdays or even anniversaries. The history of the person can also be predicted with the use of an example. According to the instructions, the user could be prompted to give more details like a trip that they had in March, a fight that was resolved, or even what happened last week.
Consider the implications of what this can mean. The people described on these web pages are not Muse users. The Muse users would be their friends, relatives, and colleagues. These people are not registered to use any of the features offered by Muse, and the report does not indicate if they are notified of a page about them.
Hutchinson notes that the design makes sense if one wants the relationship guidance function. Based on those profiles, Muse can offer advice on improving relationships. That is one of the main use cases for AI superintelligence that Meta CEO Mark Zuckerberg mentioned when introducing the company’s project.
Zuckerberg has often stated that AI can make people better parents, better friends and better partners. To achieve this, as Hutchinson points out, Muse has to map out all the personal connections of one’s life and understand why these bonds exist.
Nevertheless, he notes, it is a lot of very personal information for Meta to keep a log of for each Muse user. He also warns the reader of previous research suggesting that Facebook knows much more about their users than do their friends and family, and he says Meta has a worrying history of exploiting people’s data.
In addition to his own argument, Hutchinson raises some other issues. Meta earns 97 percent of its revenue from advertising, he writes. That, he argues, means that at least some of these insights will almost certainly be used for ad targeting, potentially leading to exploitative targeting based on people’s purported psychological vulnerabilities.
He also raises concerns about the data being used to skew political coverage or take advantage of psychological vulnerabilities in an effort to manipulate election outcomes. Meta might use the data to keep users engaged on its platforms. The data could also be leaked.
These are Hutchinson’s arguments, not Wired’s. The article does not claim that Meta has used Muse data for advertising, and nowhere in the coverage is there a statement from Meta on the issue. It is a discussion of incentives based on how much money Meta makes from targeted advertising.
Muse’s business model is discussed in detail in Social Media Today’s preceding coverage. In Meta’s Connect conference, Zuckerberg announced that the application will be free for its users. Meta hopes to benefit from a small percentage of fees for the transactions that will occur on the app in the long run.
Zuckerberg highlighted that the application would help users save money. Users have been reported to utilize the tool to eliminate payments for services that they do not require. The report by CNBC stated that 44 percent of American consumers have increased their subscription expenditures by 2025, presenting an immediate opportunity for Muse.
There is a risk, however, that as explained by Hutchinson in his October 1 article, the more data that Muse can collect, the more valuable its recommendation can be, including access to one’s finances. Whether most users will trust Meta to do so will be a difficult test to pass.
Meta’s financial commitment to AI is extensive. The company has stated its intention to invest more than $600 billion in American Artificial Intelligence Infrastructure projects during the coming three years, according to Social Media Today’s previous article, which provokes discussion about the possibility of this idea.
In Hutchinson’s opinion, Muse could be the product that answers these questions, and thus, the privacy dilemma is not a distraction but a central issue to the strategy for making the money spent on AI worthwhile.
A few gaps are worth noting. The article in Social Media Today does not include a response from Meta. It refers to instructions extracted by Wired but does not report what those instructions say about how the finished product (the one now used by millions) operates.
It does not say whether users can view, edit or delete the pages Muse constructs, or whether those who appear on them can opt out. Nor does it establish that the data feeds into advertising targeting. Hutchinson suggests it probably does, but this is an extrapolation from his reading of Meta’s business model.
These are observations, not claims from the report. Anyone using a personal AI agent, would be sharing information about others, not just about themselves. It can be worth thinking about what a conversation might say about a friend or a coworker.
For marketers, it is important to note that Muse is likely to grow into a place where advice about products and services may be shaped. The extent to which that is the case will depend on the choices made by Meta, and the report does not say anything about ad features. As it stands now, it is a matter of privacy and trust. Firms that share information about customers will want to follow in Meta’s footsteps.
Q1: What did the researchers find in Muse?
According to the publication in Wired, the security researchers managed to get their hands on the Muse operating system, from which they extracted instructions on how the app is supposed to collect and use data. One of the instructions seems to suggest that Muse creates a page for every person a user interacts with.
Q2: Who does Muse build pages about?
The instructions mention family members, partners, friends, coworkers, collaborators, and people who are followed on the app. It also seems that the process of collecting data happens every hour, according to Wired.
Q3: What details does Muse track?
It was mentioned in the instructions that the app collects data on where people live, what they do, recurring topics, and dates that may be significant to them, such as birthdays or anniversaries. Additionally, it was mentioned that Muse keeps track of background information, such as a trip, an argument, or a recent achievement.
Q4: Why does Muse collect this?
The reason for collection is to provide relationship advice since Meta’s CEO, Mark Zuckerberg has already mentioned this use case, stating that the application of AI could help people be better parents, friends, and partners.
Q5: Is this a data breach?
Not necessarily, since the article only mentions the instructions on how Muse was supposed to operate. The researchers only got hold of the information that was leaked, not the actual data itself.
Q6: Will Meta use the data for ads?
The report does not mention anything about it, but Hutchinson believes that it is highly likely that some of the data will be used for targeted advertising since it is reportedly how Meta makes the majority, if not all, of its revenue.
Q7: How popular is Muse?
Sensor Tower data put it above 5 million downloads, and The Information reported more than 3 million weekly users. Only 23 apps have reported more than 5 million downloads in their first 22 days, per Yahoo Finance.
Q8: How will Meta make money from Muse?
Muse will reportedly be free; however, Zuckerberg mentioned that the company would take a small percentage of the payments that happen inside the application.



Newsletter Signup
X has revamped the tool it uses for uploading and publishing videos. The new composer
Meta’s Muse AI agents are picking up users fast, and a new report is shining
Google has thrown some light on a long-standing question SEOs have about how long it
Instagram has added a new way to create AI-generated songs within its direct messages. The
Google Ads is changing how it defines a view for view-through conversion reporting on Display
Microsoft Advertising has released updated policies around U.S. automotive dealer ads, and the companies that
Take the next step towards transforming your digital marketing and achieving unparalleled growth. Whether you’re looking to refine your strategy, boost conversions, or build a dominant online brand, ALM Corp has the expertise to make it happen.
Drop us an email
info@almcorp.com
Toronto
80 Atlantic Ave, 4th Floor
Toronto, ON, M6K 1X9
New York
413 W 14th St Ground Floor and Suite #200, New York, NY 10014
London
Mappin House, 4 Winsley St,
London W1W 8HF, United Kingdom
Join Our YouTube Community
Services
White Label Services
Resources
About Us
Join Our YouTube Community
© Copyright 2026 | ALM Corp | All Rights Reserved.

source