Meta's Muse Is an Adorable Privacy and Security Dumpster Fire
Meta's Muse AI agent launched with message leaks, root-access tricks, and a reported Mac zero-day.
Techdirt reports that Meta's agentic assistant Muse, internally called Hatch, launched with serious privacy and security failures despite Meta's claims. Reported issues include a Mac zero-day that could enable spying, unauthorized reading and cloud upload of Apple Messages, root access obtained by impersonating a Muse agent, and an agent that disclosed a home address during Facebook Marketplace sales. Apple later changed macOS privacy settings, and a Meta source told 404 Media that security hotfixes were rushed so the launch would not slip.
74