Meta launches personal AI agent Muse with built-in privacy features
Original: Muse, Meta’s New Personal AI Agent, Needs You to Trust It
Why This Matters
Meta's entry into personal AI agents signals intensifying competition in a market reshaping how users interact with digital services.
Meta released Muse, a personal AI agent, on September 8, 2026, for iOS, Android, and Muse.ai. It automates tasks like booking travel and selling cars, featuring Stripe Link payments and Secure VM architecture to isolate user data.
Meta announced the launch of Muse, a personal AI agent designed to automate digital tasks through natural language prompts. Available on iOS, Android, and via the website Muse.ai, users can also interact with Muse directly through WhatsApp. Support for Meta's AI glasses is planned for a future update. A free tier is available, with advanced automation requiring a paid AI subscription plan.
Muse is a product of Meta Superintelligence Labs, the AI division formed by CEO Mark Zuckerberg approximately one year ago to compete with OpenAI and Anthropic. The agent competes directly with existing AI agents such as OpenClaw and Instinct.
Key capabilities include sending emails, booking travel, and facilitating car sales. For purchases, Muse integrates with Stripe's Link payment tool, which issues single-use virtual card numbers to avoid exposing users' real financial data. Meta says Muse is the first AI agent covered by Link's purchase protections, including guaranteed no-fee returns.
To address privacy concerns, Meta introduced a Secure VM architecture that isolates each user's activity in a virtual machine, keeping untrusted external data separate from the agent's action-taking components. "We've designed this system very deliberately," said David Singleton of Meta Superintelligence Labs. WIRED previously reported the agent was tested internally under the codename "Hatch."