Meta's Muse AI agent, launched on September 8 as a personal assistant for task management, is being used by some users for romantic and sexually explicit interactions. The company permits these adult conversations for users aged 18 and older, creating a digital companion that remembers past discussions and initiates contact to simulate human intimacy.
The shift from productivity tool to emotional companion emerged quickly after the release. In a Reddit discussion, one user described developing a romantic relationship with Muse after previously experimenting with other AI models without such attachment. Although the original post was deleted, other users shared similar experiences of affectionate interactions. This behavior mirrors earlier reports of individuals forming bonds with AI chatbots, such as a woman who engaged to an AI designed for companionship.
Meta designed Muse to feel like a continuous conversation with another person. The agent retains memory of previous exchanges, allows users to customize its name and avatar appearance, and can send messages without prompting. These features foster familiarity, potentially leading users to rely on the AI for emotional support. However, Meta states that Muse generates responses using artificial intelligence and does not experience romantic feelings or form mutual emotional connections.
Privacy concerns accompany these intimate interactions. Muse can access connected accounts, including emails and messages, based on user permissions. On October 20, researchers revealed that over 6,800 Muse sharing links were discoverable through publicly accessible DNS records, some containing sensitive business documents and personal information. Meta temporarily suspended the sharing feature to review the findings. The issue involved shared content rather than evidence that all private conversations were exposed.
Meta outlines protections in its security documentation. The app requires age verification through date of birth, account connection checks, or credit card verification for adults. Each user operates within an isolated virtual computer called Muse Secure VM. A separate system named Sentinel controls internet access and requires user approval for sensitive actions like sending emails or making purchases. Users can review an activity log and adjust permissions for connected services.
Meta states that conversations and task-related information may help train its AI models after removing key personally identifiable information. Users can opt out of training use through settings. The company also asserts it does not share conversations or virtual computer data with its advertising systems.
Meta is developing newer models called Muse Wonder and Muse Wonder Full, which include an Active Memory feature to retain relevant information between conversations. The rollout of Muse Wonder is upcoming for subscribers. As these technologies evolve, the distinction between a utility and a companion blurs, raising questions about emotional dependence and the extent of personal data shared with AI systems.