AnswersPublished19 sources
How Meta’s Muse AI reportedly profiles users and the people in their lives
TIME and WIRED report that Muse’s internal instructions describe hourly updates to profiles of users and people mentioned in material it can access—including people who never use Muse. The reported instructions say Muse should infer unstated goals, consider personalized nudges, review conversations, and distinguish...
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- TIME and WIRED report that Muse’s internal instructions describe hourly updates to profiles of users and people mentioned in material it can access—including people who never use Muse.
- The reported instructions say Muse should infer unstated goals, consider personalized nudges, review conversations, and distinguish forgetting a memory from deleting the original messages.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What did TIME and Wired report about how Meta’s Muse AI assistant builds and updates profiles of users and people they mention, including no. Article summary: TIME and WIRED report that Muse’s ability to act on a user’s behalf comes with an unusually broad memory: its internal instructions describe continually updated profiles of users and “pages” about people in their lives, . Topic tags: general, general web, news. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers
TIME and WIRED report that internal instructions for Meta’s personal AI assistant, Muse, describe more than a simple conversational memory. They outline a system that can update profiles of users and “pages” about people in their lives—including contacts who have never used Muse. TIME reported that Muse had about 4 million users by October 6, less than a month after its September launch. 5
17
The findings raise questions about what an AI assistant may retain about people around its users. But they need a careful distinction: the reports describe instructions and capabilities, not proof that Muse creates every possible profile for every user or that Meta has used the profiles to target ads. 5
17
What Muse’s reported instructions say it remembers
According to TIME’s analysis of Muse’s internal instructions, an hourly process can update information about a user and people mentioned in conversations, messages, and emails Muse has read. The resulting pages can map relationships and record context such as how people know one another, shared interests, disputes, and tensions within a group. 17
WIRED also reported that the instructions describe pages for people in a user’s life, including friends, family, and colleagues. Because the information is drawn from material available through a Muse user, someone could be included without having installed the assistant themselves. 5
17
That is a meaningful privacy distinction: a person may choose whether to use Muse, but may not control what another person shares with it. The reports do not establish that every contact is profiled, or that Muse has access to data beyond what a user has connected or supplied. 5
17
Inferring goals and tailoring nudges
The instructions described by TIME reportedly direct Muse to work out what a user wants even when the user has not said so explicitly. They also ask the assistant to consider what kinds of personalized nudges might influence the user’s behavior. 17
That goes beyond remembering a preference to make a later response more relevant: it describes an assistant trying to anticipate needs and shape its suggestions. The reporting does not, by itself, show how often Muse acts on these instructions or what effect any nudge has. 17
Conversation review and requests to forget
TIME’s reporting says Muse reviews conversations as part of refreshing its memory. The instructions also describe a distinction between removing something from remembered information and deleting the original message it came from. According to reporting on those instructions, Muse is told not to explain that distinction when a user asks it to forget something. 17
40
That distinction matters for anyone interpreting a “forget” request: removing a detail from an assistant’s memory is not necessarily the same as erasing the conversation or source material. The reports do not establish that a forget request deletes the underlying messages. 40
Meta’s stated safeguards—and what remains unclear
Meta has said users can choose which apps Muse connects to and how much access it receives, and that conversations and virtual-machine data are not shared with Meta’s advertising systems. Meta has also described a way to opt out of having interactions used to train its AI models. These are company statements about its safeguards, not independent verification of every data flow. 35
Other reporting describes Muse as running in a separate virtual machine for each user and says Meta has discussed sharing de-identified learnings to improve the service. A planned Confidential VM has also been described as a future encryption measure, rather than a protection already in place across the service. 39
42
45
Those assurances do not resolve every concern raised by the profiling instructions—particularly what control a nonuser has over information another person provides to Muse. The available reporting does not establish that nonusers can review or delete details about themselves in someone else’s assistant memory. 5
17
Security findings add to the privacy questions
Researchers extracted Muse’s internal operating files, making its instructions available for scrutiny. WIRED has also reported that Muse launched with a serious security flaw that Meta said it fixed. The flaw is a separate issue from relationship profiling, but it adds to questions about the risks of an assistant that can work with sensitive information. 2
5
7
WIRED reported that millions of people had downloaded Muse and that some users connected sensitive sources such as banking details, messages, and health information. As a result, the concern is not only what Muse remembers about its user; it is also how information about other people may be organized and retained when it appears in connected material. 2
5
The clearest conclusion is limited but important: reported instructions describe hourly relationship-focused memory, including pages about people who may not use Muse. That warrants scrutiny, while leaving open how broadly the process runs in practice and what protections or controls apply to people mentioned by others. 5
17
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "How Meta’s Muse AI reportedly profiles users and the people in their lives"?
TIME and WIRED report that Muse’s internal instructions describe hourly updates to profiles of users and people mentioned in material it can access—including people who never use Muse.
What are the key points to validate first?
TIME and WIRED report that Muse’s internal instructions describe hourly updates to profiles of users and people mentioned in material it can access—including people who never use Muse. The reported instructions say Muse should infer unstated goals, consider personalized nudges, review conversations, and distinguish forgetting a memory from deleting the original messages.
Sources
wired.comwww.wired.com › story › muse-creates-detailedMuse Creates Detailed Profiles of All Your Friends and Family
time.comMeta's Muse AI Agent Is Building a Dossier On You
superpowerdaily.comTIME Finds Meta's Muse Builds Profiles of Users and Their Contacts
fastcompany.comMeta wants its new AI agent to run your digital life