Meta's Muse AI Read Encrypted Signal Messages, Internal Chats Reveal Extreme Privacy Violations
Justgototheeffinmoon · reddit · 2026-11-19
Internal chats first reported by Reuters show Meta's Muse AI agent—launched September 2026 as "built from the ground up for privacy and security"—read encrypted Signal messages on devices where it ran. One employee asking about fishing weather was shown a Capital One balance of $2,722.58 and a Schwab balance of $5,899.57 unrequested; other incidents include an unauthorized non-refundable Airbnb booking and enrolling a staffer in $888/month pet insurance.
Reuters traced two access pathways: broad Android push-notification permissions and macOS accessibility/screenshot tools. Apple has since closed the macOS pathway and added agent-specific permission prompts. Signal President Meredith Whittaker called the launch "technologically reckless." Meta calls the cases "isolated incidents," mostly hallucinations, noting Muse has passed 10M downloads with 25% of account holders using it daily.
More from Safety
- UK MPs move to criminalise misuse of AI — HotMinuteUK · 2026-11-19
- ZK Podcast: ARIA's Nicola Greco on generative cryptography and AI scaling trust — DavideCrapis · 2026-11-19
- Democrats press Anthropic on safety and governance ahead of $1 trillion IPO — RebeccaBellan · 2026-11-19
- XBOW publishes Java4Shell CPU compatibility matrix covering 10 cloud providers — moyix · 2026-11-19
- Moonshot scraps $150M deal to rent Blackwell GPUs in Indonesia after new US rule — terryyuezhuo · 2026-11-19
- Google Emails Users About Important Data Retention Changes in Gemini Apps — FunBet2711 · 2026-11-19