Meta’s AI agent Muse is under investigation after reports reveal it builds comprehensive profiles of individuals in a user’s social circle, raising new privacy concerns amid tightening AI regulation.
Meta’s personal AI agent Muse is drawing fresh scrutiny after reports that it builds detailed profiles of the people in a user’s life, including relatives, friends, partners and colleagues who have never signed up for the service. According to Wired, internal files and instructions indicate that the system creates a separate page for each relevant person and then updates those records over time using material drawn from conversations and its memory.
The material suggests that Muse does more than store names or basic contact details. It is designed to assemble a living picture of relationships, including where someone lives, what they do for work, how they know the user, birthdays, anniversaries, recent conversations, shared experiences, resolved disputes, common goals and even the preferred way to communicate with them. In practice, that means Muse can build a social map around the user without the other person ever interacting with Meta’s AI directly.
The disclosure lands at a sensitive moment for the wider AI industry. Apple recently signalled tighter controls in macOS over concerns that more autonomous agents could gain access to sensitive data on personal computers. Other firms are also moving quickly to harden their systems: Nvidia has launched an Open Agent Safety Platform to isolate and monitor agent behaviour, reflecting growing unease about tools that can act with broad permissions across apps and services.
Muse has already faced separate criticism over privacy. TechRadar reported that a Surfshark analysis found Meta’s new chatbot among the most data-hungry AI tools tested, collecting 31 of 35 possible data types and trailing only Meta AI itself. The same report said the service can access highly sensitive information, including calendar entries, emails and bank details, while critics warned that such data could be used for advertising or model training.
Adoption of these tools remains uneven. Axios reported that, despite heavy investment across the industry, most people are still reluctant to give AI agents deep access to their digital lives. That hesitation may prove especially relevant for a system like Muse, whose value depends on unusually broad access to personal data. Meta has also been accused of making its assistant too intrusive on Mac and iPhone, with Tom’s Hardware reporting claims that it referenced confidential messages without permission; Meta says message access is opt-in and requires both Full Disk Access and the Messages connector.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
