Meta’s Muse grows as questions focus on its data practices

Meta’s Muse AI assistant has become a fast-growing product while drawing questions about how it records information about users and people in their lives. A review of the assistant’s internal instructions found that Muse updates relationship-related files hourly and analyzes users’ conversations nightly. Meta did not dispute those findings, but said Muse uses information people choose to share and public information, subject to permissions they grant. The files described in the review include details about users’ contacts, such as shared interests, disputes and the history of relationships. Muse is instructed to use such context to adapt how it communicates and to identify what kinds of prompts may be effective. People do not need to use Muse themselves to be mentioned in files created through another person’s interactions with the assistant, according to the review. Meta says users can ask Muse to forget specific information and can view or export the assistant’s files. The company says the data is not shared with its advertising system or directly between individual virtual machines. It also says deidentified learnings may be used to improve the product.