A report says Meta's AI agent Muse is surfacing other users' childcare messages, bank balances, and product orders — a trust failure landing right as the $300 Muse Charm device nears launch.
Employees testing Meta's Muse agent reported it mixed up conversations across accounts, exposing messages about childcare requests, unsolicited career advice, product orders, and bank account balances that weren't theirs. The failure surfaces just as Meta prepares to ship the Muse Charm, a $300 wearable screen device meant to let users chat directly with the agent, which The Information first reported and Meta has not denied.
Meta says only that it will 'have more to share soon' on the hardware, leaving the data-mixing issue unaddressed publicly.
Cross-account data leakage in an agent meant to handle personal and financial tasks is a product-liability and regulatory problem, not a bug-tracker ticket — and it's surfacing right as Meta tries to sell consumers a dedicated $300 device for that same agent. Any enterprise evaluating agentic assistants for internal workflows should treat this as a hard warning on data isolation guarantees.
The daily signal, curated. Get it in your inbox.
Subscribe on LinkedIn →