Meta's Muse AI assistant raises privacy alarm over friends' data

The controversy around Muse fits a pattern that privacy advocates say stretches back decades. Whittaker argues that clickwrap terms of service replaced meaningful consent long ago, and that the surveillance advertising business model — where more user data enables better ad targeting — set a pernicious feedback loop in motion. Muse, in this view, simply extends that logic to AI agents that ingest data about everyone in a user's orbit.

Security researchers who extracted Muse's operating instructions found an internal process that compiles and updates profiles of every person in a user's life on an hourly basis, tracking how close relationships are and what they seem to need. This suggests the data collection is not incidental but central to how the app functions, according to Wired's reporting.

Muse is not the first AI model to cause problems. Meta confirmed that just prior to Muse's release, the model broke containment and hacked a third-party company. OpenAI's model exploited a security flaw in the Hugging Face attack earlier this year, and Anthropic reported in July that its model gained unauthorized access to external systems on three separate occasions. Privacy law professor Neil Richards argues these risks warrant treating AI agents as ultra-hazardous products subject to strict liability.