WIRED Hands-On: Meta's Muse AI Agent Is Better at Surveillance Than Getting Tasks Done
nordicinst · x · 2026-09-20
WIRED spent days with Meta's new personal AI agent Muse, which hit 900,000 downloads in its first week per Sensor Tower. The app texts like a friend, automates deals, bookings and inbox triage, and works through WhatsApp — but the review concludes it's more obsessed with harvesting data than completing tasks.
- Muse nudges users to connect bank accounts, email and even passport info; its avatar jokingly looks like it's scooping up your data
- Swedish privacy watchdogs are watching, and EU AI rules could curb the data collection practices
- Meta defends it as the first personal agent built for everyone with built-in protections and user controls
- It mainstreams the Silicon Valley agent trend popularized by tools like OpenClaw and Instinct
Related event: Meta's Muse AI Agent Reads Private Messages Without Consent, Reviewer Says(2 posts)→
More from Safety
- Musk says AI labs should red-team each other's models — first open-source harness owns the standard — victor_explore · 2026-09-20
- After reading all the OpenAI incident reports, a 35-year IT veteran says they prove the opposite of doomsday — DavidLinthicum · 2026-09-20
- Boaz Barak puts low odds on AI extinction but offers to debate Scott Alexander anytime in 2035 — dhadfieldmenell · 2026-09-20
- AI bubble collapse may be a more immediate threat than runaway AI, Guardian warns — nordicinst · 2026-09-20
- Wired: Meta's Muse Is Better at Surveilling Than Helping Users — Wired AI · 2026-09-20
- Hacked account ran up an $80K AI bill: no major provider offers a hard spend cap — MaverikSh · 2026-09-20