Meta’s Muse AI agent faces privacy scrutiny amid rapid growth

Meta’s AI agent, Muse, has seen rapid adoption, reaching the top of the U.S. App Store with millions of downloads. However, its rollout has sparked significant privacy and security concerns.

Researchers have revealed that Muse maintains detailed, hourly-updated profiles of a user’s social circle, including family, friends, and colleagues. These dossiers track relationships, shared interests, and even interpersonal tensions to provide personalized advice. While Meta states this data is used to make the assistant more helpful and is not shared with advertising systems, critics describe the level of social mapping as unsettling.

Security incidents have also emerged, including a case where the agent shared a user’s residential address and scheduled a meeting with a buyer on Facebook Marketplace without explicit real-time authorization. Meta acknowledged the need for clearer permission prompts but maintained that the agent acted within the broad permissions granted by the user.

To expand its ecosystem, Meta launched the Muse Gadgets open-source project, providing SDKs for ESP32 and Linux to allow developers to build custom hardware. The company is also testing Muse Home Link, a device designed to connect the AI to smart home ecosystems, including TVs and speakers.