{"slug": "meta-launches-muse-ai-agent-what-it-can-do-and-how-it-handles-your-data", "title": "Meta Launches Muse AI Agent: What It Can Do and How It Handles Your Data", "summary": "Meta launched Muse, a personal AI agent powered by its Muse Spark large language model, on September 9, 2026, initially available in the United States. The agent can handle tasks such as sending emails, booking travel, and making purchases, and Meta claims user data will not be shared with advertisers. Muse is free to use, with paid plans Muse Power at $20 per month and Muse Maximum at $100 per month, and is rolling out in WhatsApp with iOS, Android, and web availability.", "body_md": "**September 9, 2026**, (Inside AI) — Meta launched a new personal AI agent called **Muse** on Tuesday, designed to handle everyday tasks like sending emails, booking travel, and making purchases. The agent is powered by Meta's large language model **Muse Spark** and is initially available only in the **United States**.\n\nMuse connects to a user's email, calendar, payments apps, health and fitness apps, smart home devices, e-commerce platforms, and dining apps. It can fill out forms, create plans, turn recipes into grocery lists, send party invitations, and complete checkout purchases. Meta said it plans to add more connectors over time.\n\nThe launch puts Meta into an already crowded agentic AI market. But Meta's edge may not be the agent itself. The parent company of **Instagram** and **WhatsApp** has a distribution network that could embed Muse deeply into billions of digital lives. Still, its biggest hurdle may be trust. Meta's history on privacy and data use could make users hesitant to hand over more personal information.\n\n## Muse Spark's Mixed Benchmark Record\n\nMuse Spark is the first AI model released in **April 2026** by **Meta Superintelligence Labs**, the costly team Meta assembled to catch up in the AI race. Independent evaluations show Muse Spark catching up to top models from **Google**, **OpenAI**, and **Anthropic** in language and visual understanding.\n\nBut it lags in coding and abstract reasoning. That gap matters for an agent expected to execute complex, multi-step tasks reliably. Meta did not disclose specific benchmark scores in its announcement.\n\n## Security Architecture and Data Isolation Claims\n\nMuse runs on a dedicated, secure virtual machine called **Muse Secure VM**, which has its own browser. Meta says a separate **Sentinel** agent runs on the same VM but is kept apart at the system level. That separation means Muse cannot access user passwords or payment methods directly.\n\nMeta said user conversations and data shared with Muse will not be shared with Meta advertisers. Users can also choose which apps and services to connect. If a platform lacks a public **API**, Muse can set up a connection using credentials the user provides. When no API is available, Muse can access the service via the browser instead.\n\nUsers can customize Muse by giving it a name, picking an avatar, and configuring how it communicates. The agent continues working after the user leaves the app. It learns from user chats over time to make unprompted suggestions.\n\nThe agent is free to use, but paid plans kick in as usage grows. **Muse Power** costs **$20** per month, and **Muse Maximum** costs **$100** per month. A usage meter in the app shows what percentage of free usage remains.\n\nMuse is rolling out in **WhatsApp**, with a **Meta AI glasses** integration in the works. It is also available through standalone **iOS** and **Android** apps, and on the web at [muse.ai](https://muse.ai).\n\nMeta's agentic push follows its recent launches of **Muse Voice Transcribe** with support for five Indian languages and **Muse Glimmer**, which reignited the open-weight versus closed AI debate. The company has not said when Muse will expand beyond the US.", "url": "https://wpnews.pro/news/meta-launches-muse-ai-agent-what-it-can-do-and-how-it-handles-your-data", "canonical_source": "https://insideai.news/news/agentic-ai/meta-muse-ai-agent/9976/", "published_at": "2026-09-09 07:24:15+00:00", "updated_at": "2026-09-09 07:29:08.376326+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-agents", "ai-products", "ai-ethics"], "entities": ["Meta", "Muse", "Muse Spark", "Meta Superintelligence Labs", "Google", "OpenAI", "Anthropic", "WhatsApp"], "alternates": {"html": "https://wpnews.pro/news/meta-launches-muse-ai-agent-what-it-can-do-and-how-it-handles-your-data", "markdown": "https://wpnews.pro/news/meta-launches-muse-ai-agent-what-it-can-do-and-how-it-handles-your-data.md", "text": "https://wpnews.pro/news/meta-launches-muse-ai-agent-what-it-can-do-and-how-it-handles-your-data.txt", "jsonld": "https://wpnews.pro/news/meta-launches-muse-ai-agent-what-it-can-do-and-how-it-handles-your-data.jsonld"}}