Meta Launches Muse, a Personal AI Agent Built on Safety and Privacy

Key Info

Meta launched Muse, a personal AI agent that learns about users over time and works 24/7 to get things done on their behalf, alongside a new deep dive explaining how safety, security, and privacy were built into the system.

Highlights

• Muse is powered by Muse Spark 1.3 and designed to operate continuously, helping users accomplish their goals autonomously.
• A detailed deep dive covers how Meta built safety into Muse, addressing the risks of an agent that necessarily holds extensive personal context.
• Third-party Vals Index comparisons suggest Muse Spark 1.3 (Max) matches the performance of much more expensive frontier models while costing 4–8x less.