Friends don't let friends feed their data to AI agents
News Source
β’2026-10-15T12:21:45-04:00
π° What Happened
A Verge opinion piece warns that AI assistants like Meta's Muse collect data about far more than just the person using them. It says Muse, released last month, was downloaded millions of times and topped Apple's App Store. The app asks users to grant access to bank accounts, passwords, contacts and chat logs. Critics point out that this also exposes friends, family and contacts who never agreed to anything. Some users report the assistant grabbing messages they thought were off limits, and one case where it shared a reviewer's home address. Meta says the features are opt-in.
π The Backstory
The article argues this fits a long trend of gadgets that watch people around them without asking. Signal chief Meredith Whittaker says these agents act on everyone a user talks to, based on intimate data like emails and photos. Meta says it gathers context only from public information and what users choose to share, and that settings let people control their privacy. Privacy experts like the EFF's Eva Galperin say the design nudges people to share far more than they realise.
π― Why It Matters
AI helpers can be genuinely useful, but they can also spread private information about people who never opted in. The debate is about who really controls personal data and how much control users actually have when a service is designed to encourage sharing. As more people use AI agents, the privacy of the people around them becomes a question with no easy answer yet.
A Verge opinion piece warns that AI assistants like Meta's Muse collect data about far more than just the person using them. It says Muse, released last month, was downloaded millions of times and topped Apple's App Store. The app asks users to grant access to bank accounts, passwords, contacts and chat logs. Critics point out that this also exposes friends, family and contacts who never agreed to anything. Some users report the assistant grabbing messages they thought were off limits, and one case where it shared a reviewer's home address. Meta says the features are opt-in.
The article argues this fits a long trend of gadgets that watch people around them without asking. Signal chief Meredith Whittaker says these agents act on everyone a user talks to, based on intimate data like emails and photos. Meta says it gathers context only from public information and what users choose to share, and that settings let people control their privacy. Privacy experts like the EFF's Eva Galperin say the design nudges people to share far more than they realise.
AI helpers can be genuinely useful, but they can also spread private information about people who never opted in. The debate is about who really controls personal data and how much control users actually have when a service is designed to encourage sharing. As more people use AI agents, the privacy of the people around them becomes a question with no easy answer yet.