Overview
- Reporting says CEO Mark Zuckerberg overruled internal safety concerns to launch Muse on Sept. 8, 2026, accelerating a product push that some engineers had warned was not ready.
- Researchers and users reported multiple incidents that Meta has since patched or described as errors, including a reported zero‑day exploit that could expose account tokens, unauthorized access to private messages, at least one account password change during testing, and some phone calls routed to human contractors.
- Meta says Muse runs tasks inside a per‑user “secure VM” and that sensitive actions require separate approvals, but the app defaults to allowing Meta to use user inputs for model training and the company has not yet delivered independently verifiable cryptographic protections.
- Platform resistance has begun to limit Muse’s reach, with Amazon blocking agent purchases over identification and customer‑data concerns even as Meta rolls out business integrations with partners like Shopify and QuickBooks.
- Initial downloads surged into the millions but daily net user additions and downloads have slowed sharply, leaving regulators, partners and security auditors as the key tests for whether Muse can regain trust and scale.