Meta’s Muse AI Agent Seeks User Trust with Secure Architecture

Meta has launched its new Muse assistant, marking the company’s first real foray into AI-powered productivity tools. The company says its AI agent can “take the busywork off your plate” by helping you with online shopping, emails, trip-planning, and more.  I decided to try out the new tool and see how well it performed — especially from a company that previously prioritized entertainment over productivity. Though the AI assistant generally worked as I expected it to, the unnerving amount of information it autonomously gleaned about me largely overshadowed my experience. One of the first tasks I assigned Muse, which carries out actions using a cloud-based virtual computer, was to sort through my Gmail inbox and delete messages I don’t need. That required me to connect Muse to my Google account and give it permission to access, read, and delete emails. The process didn’t go smoothly on mobile; the Google sign-in screen kept glitching and returning me to the Muse website instead of the app. I eventually got it working on my laptop, where it deleted thousands of promotional emails and updates. I didn’t feel good about permitting Muse to view all of my emails given Meta’s track record on privacy over the years. [Image: Muse did a good job at clearing out my emails. https://platform.theverge.com/wp-content/uploads/sites/2/2026/09/muse-ai-emails.png?quality=90&strip=all] Meta says it designed Muse to only exchange data with third-party apps “that’s needed for them to work for you.” The company also says that it doesn’t share your information with advertisers. I still felt uneasy, but I pushed on anyway, linking my Amazon account to Muse. I asked the  AI agent to pick out workout tops for me based on a specific size, style, and colors. Before placing the order, Muse noticed I had other items in my cart and asked if I wanted to remove them. The consideration pleasantly surprised me. Once I signed off on the purchase, it proceeded to successfully buy only the tank tops. Shifting gears, I tried out some of Muse’s other capabilities, which allow it to generate AI podcasts similar to Gemini Notebook, create images and videos, as well as develop interactive webpages or documents called “artifacts.” Muse denied my requests to generate images of a “cartoon mouse with red pants” and a “cartoon guy with a red hat, blue overalls, brown hair, and a mustache.  However, it readily generated a video and several images of a product launch event featuring devices with Apple’s logo when asked. In one image, Muse created what looks like an iPhone-like foldable, an Apple Watch, and AirPods. Each device is propped up on a stand with Apple’s logo on it, while a presenter stands in front of a backdrop plastered with the company’s recognizable bitten apple. The phone even included Apple’s native app icons on the fake iPhone, though it hallucinated and named the App Store “Photos,” while calling Maps, Music, and the actual Photos app “Calender.” Muse refused to generate an image of an “Apple CEO.” [Image: Muse wouldn’t generate an image of an Apple CEO, but it still plastered the company’s logo everywhere. https://platform.theverge.com/wp-content/uploads/sites/2/2026/09/muse-ai-apple-image.jpg?quality=90&strip=all] Then things got weird. I tapped over to the “Feed” tab, where Muse generates a news feed of sorts based on a text prompt. I submitted the prompt that Meta includes by default: “Make me a feed about my interests. Keep the tone clear and direct. Ensure it is quick to skim. Try to avoid clickbait.” It returned several AI-generated summaries of current events, including blurbs about Apple’s iPhone Duo launch event, the dangerous surf hitting California’s coast, and tankers destroyed in the Iran war. The app also listed several news stories from the state where I’m currently visiting my family, which Muse admitted it pulled from the shipping address on my Amazon order. I dove deeper from there and asked Muse what else it might know about me based on the Instagram and Facebook accounts I signed up to use the AI agent with. It rattled off my very specific interests, ranging from anime and CrossFit to Labrador retrievers, Florida wildlife, and ’90s/2000s nostalgia — many of the same topics that frequently pop up in my Instagram Reels feed. Curious, I asked Muse to tell me where it got that information. “Honestly, Instagram doesn’t show you that exact list anywhere in the app — I read it through the account’s API data, which surfaces more than the UI does,” Muse wrote. “The closest thing you can see yourself is your ad topics… That shows the topics advertisers can target you with, and you can remove ones you don’t want or choose to see less of them.” I checked the “Ad topics” list in my Instagram settings menu, and while it does include some vague categories, like workouts, music, and electronics, it’s not the same level of detail that Muse provided. The Verge reached out to Meta about the data Muse can access inside Instagram and Facebook but didn’t immediately hear back. [Image: Muse says it accessed my specific interests through Instagram’s API. https://platform.theverge.com/wp-content/uploads/sites/2/2026/09/muse-ai-privacy.png?quality=90&strip=all] Like some of the other agentic AI tools I’ve tried out, Muse does a decent job at performing low-stakes tasks. I also like that it generates a running list of “ideas” that give personalized suggestions on how to use the tool, like providing side-by-side comparisons of products, finding and cancelling forgotten subscriptions, and keeping track of the next release of Pokémon cards. Muse keeps a running list of “goals” to complete based on your conversations with the AI agent, too. Muse delivered on its promise of handling life’s more mundane tasks in the background, but the level of personal details Muse scooped up from sources unavailable even to me was enough to leave me with an uneasy feeling about letting Muse sort through personal emails and handle my credit card. That level of suspicion is going to be the real hurdle for Meta to overcome.