Meta tackles agentic AI with the launch of Muse

Meta launched Muse, an agentic AI model powered by Muse Spark that can draft emails, find deals, and complete other tasks on behalf of users, with access initially opening to iOS and Android users in the US. The system runs in a private virtual machine for privacy, with a separate agent called Sentinel ensuring actions are approved, and a Muse Confidential VM update arriving later this year to prevent even Meta from seeing user data.

Affiliate links on Android Authority may earn us a commission. Learn more. https://www.androidauthority.com/external-links/ Meta tackles agentic AI with the launch of Muse Sep 8, 2026 — 6:45 PM ET - Meta’s latest AI system is Muse, an agentic model powered by Muse Spark designed to complete tasks on your behalf. - Muse offers privacy by running within a private virtual machine. - Access is initially opening to iOS and Android users in the US. Meta has been taking strides to step up its AI game in recent months, launching Muse Glimmer https://www.androidauthority.com/meta-muse-glimmer-3696698/ and Muse Spark https://www.androidauthority.com/meta-unveils-muse-spark-ai-3656168/ earlier this year. Now it’s time for the company to make its big agentic push with the introduction of just straight-up Muse https://www.androidauthority.com/meta-muse-agentic-ai-report-3708619/ . Under the hood, Muse is powered by Muse Spark, and as Meta tells it https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/ , Muse it all about making it easier to get some work done. That means that users can present Muse with requests for anything from drafting and sending an email, to helping them find a deal on a new purchase, and the system will come up with its own plan of attack, and then move forward on actually doing the job once authorized. A separate agent called Sentinel helps make sure that Muse isn’t going online and doing anything you haven’t approved. Obviously, getting major tasks done like that is going to require you to put a lot of trust in Muse and authorize it to access your accounts, so Meta goes to some length explaining the security steps it’s taking https://research.meta.ai/blog/security-and-safety-for-ai-agents-our-approach-with-muse in order to keep your data safe, like corralling it in a personalized virtual machine. Then your interactions with that virtual box take place over encrypted connections to Meta’s apps on your devices. That security is already scheduled to receive an update, with Muse Confidential VM arriving later this year to make it so that even Meta itself can’t see your data. Like similar agents, Muse has a memory, and is supposed to learn your preferences as you express them. While all this sound fine, it’s the also the same sort of thing that all the other big AI players are doing, and it remains to be seen if Meta has the juice to make Muse the first pick users reach for when they’re choosing an AI to work on their behalf. If you’re interested in giving Muse a try, you can sign up at Muse.ai https://muse.ai/ for access. Availability is getting started in the US, with support for both iOS and Android, Meta says that Muse on smart glasses https://www.androidauthority.com/new-meta-smart-glasses-3680449/ will be coming soon. Thank you for being part of our community. Read our Comment Policy https://www.androidauthority.com/android-authority-comment-policy/ before posting.