The Social Graph as a Weapon: Meta's Muse and the Algorithmic Shadow-Profile
AI-generated image · US National Wire
Under the guise of a helpful personal assistant, Meta is architecting a system that systematically maps your entire social circle without their consent.
Opinion: Meta is not simply building a productivity tool; they are weaponizing the social graph. By deploying Muse, Meta is creating a permanent, algorithmic shadow-profile of every person in a user's orbit, transforming intimate human connections into structured data points for a machine to analyze.
As Wired first reported, researchers have extracted the internal operating instructions for Muse, Meta's viral AI agent. These files reveal a chilling directive: Muse is designed to create "a page for every person in the user's life." This is not a passive memory feature; it is an hourly process of compilation targeting family, partners, friends, colleagues, "collaborators," and anyone the user follows.
These profiles are designed to evolve. Wired reports that while a page may start "sparse," it is intended to be filled over time with specific sections: Facts, History, The relationship, In common, Open threads, and Strengthening. The system is instructed to track granular details, including where people live, their occupations, recurring "threads"—such as shared savings goals or apartment moves—and significant dates like birthdays and anniversaries.
Most invasive is the focus on the emotional architecture of these relationships. The instructions direct Muse to record how close two people are, what the relationship is built on, how they interact, and what the relationship "seems to need right now." The "Strengthening" section even suggests ways to manipulate these bonds, providing the user with "a reason to call" or "something they said to circle back on."
Karan Joshi, an independent AI safety and security researcher who extracted these prompts, told Wired that Meta is attempting to "know you like a friend," a prospect he described as "pretty creepy." This sentiment is echoed by Carissa Véliz, an associate professor at Oxford’s Institute for Ethics in AI, who told Wired that the danger lies not just in explicit data, but in what AI can infer or piece together from disparate sources.
Meta spokesperson Daniel Roberts told Wired that Muse relies on public information and data users choose to share to provide context, such as remembering a specific plumber or a spouse's favorite flowers. Meta further claims that each user has a dedicated virtual machine to store data, and that users can wipe memories or audit the agent's activity.
However, the systemic goal remains clear. Miranda Bogen, director of the Center for Democracy and Technology’s AI Governance Lab, noted to Wired that Muse places a heavier emphasis on personal contacts and relationships than its competitors. Bogen warns that these tools actively solicit users to plug in their entire lives—including financial institutions, calendars, and emails—to function.
By encouraging users to integrate every facet of their existence, Meta is not just assisting the user; it is mapping the invisible web of human connection to feed its algorithmic engine.