AIToday
Large Language ModelsAI Safety & AlignmentAI Business & IndustryITmedia AI+Published: Sep 9, 2026, 10:00 JST2 min read

Meta launches 'Muse' personal AI agent in US

Meta launches 'Muse' personal AI agent in US

3 Key Points

  1. What happened

    Meta announced Muse on September 8, offering iOS, Android, and web versions in the US, with free basic use and a subscription plan for heavier users.

  2. Why it matters

    Muse, powered by the 'Muse Spark 1.3' model, moves beyond chatbots by autonomously completing tasks like sending emails and making travel bookings, operating across its app and WhatsApp.

  3. What to watch

    Safety is a key test, as Meta acknowledges prompt injection remains unsolved and Muse may make mistakes. The company plans a 'Muse Confidential VM' by year-end.

WHO IT HITSIndividual consumers in the US who want an AI agent to handle personal tasks like email, shopping, and forms are the direct users. Businesses selling goods or services through these automations, such as retailers and travel providers, may see new types of AI-driven customer interactions.

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Context & Analysis

Meta's announcement of Muse represents a significant step into the agentic AI space, moving from answering questions to performing actions. The company highlights its focus on accessibility, claiming it is the first personal AI agent built for billions of users, and it leverages its ownership of platforms like WhatsApp and Instagram for integration. The underlying model, 'Muse Spark 1.3', is described as their most powerful to date, and a vision for 'personal superintelligence' has been previously outlined by Mark Zuckerberg.

The technical architecture shows a strong emphasis on safety and user control, with a dedicated secure virtual machine and a separate 'Sentinel' agent to oversee actions. Acknowledging unresolved industry-wide issues like prompt injection, Meta has implemented multi-layered defenses and a bug bounty program. The success of Muse may hinge on how well it balances user autonomy with safety, and whether it can execute real-world tasks reliably enough to gain consumer trust.

FAQ
What can 'Muse' do?
Muse can autonomously perform tasks like sending emails, booking travel, filling out web forms, and negotiating on the user's behalf. It can also handle multi-step goals, such as converting saved recipe videos into a shopping list or suggesting dinner menus based on a friend's dietary restrictions.
How does 'Muse' handle payments and security?
For payments, Muse uses Stripe's digital wallet 'Link', which issues disposable card numbers so actual card details are not shared. It runs on a dedicated virtual machine ('Muse Secure VM') and uses a separate agent 'Sentinel' to approve all operations, preventing access from other agents.
Is 'Muse' free to use?
Meta says basic use of Muse is free. A subscription plan will be offered for users who want to use it more extensively.

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Meta launches Muse AI agent, stresses securitySiliconANGLE AI · 1h ago
  • OpenAI launches ChatGPT Images 2.5 with SketchSiliconANGLE AI · 1h ago
  • OpenAI claims AI solved Navier-Stokes; mathematician cries foulITmedia AI+ · 1h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleStartup Calif alerts Tencent to WeChat hacking worm