Meta's Muse AI Handles Tasks but Raises Privacy Concerns in The Verge Test
Meta's Muse AI works on tasks like email and shopping, but a Verge test found it intrusive in surfacing personal data.
The Verge reporter first asked Muse, which carries out actions using a cloud-based virtual computer, to sort through a Gmail inbox and delete unneeded messages. That required connecting Muse to a Google account and giving it permission to access, read and delete emails. The process did not go smoothly on mobile; the Google sign-in screen kept glitching and returning the reporter to the Muse website instead of the app. The reporter eventually got it working on a laptop, where it deleted thousands of promotional emails and updates. The reporter said allowing Muse to view all emails felt uncomfortable given Meta's privacy record.
Meta says Muse was designed to exchange data with third-party apps only "that's needed for them to work for you," and that it does not share user information with advertisers. The reporter linked an Amazon account and asked Muse to pick out workout tops in a specific size, style and colors. Before placing the order, Muse noticed other items in the cart and asked whether the reporter wanted to remove them. The reporter called the consideration a pleasant surprise and said Muse successfully bought only the tank tops.
The reporter also tested Muse's other capabilities, which include generating AI podcasts similar to Gemini Notebook, creating images and videos, and developing interactive webpages or documents called "artifacts." Muse denied 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." It readily generated a video and several images of a product launch event featuring devices with Apple's logo. In one image, Muse produced what looked like an iPhone-like foldable, an Apple Watch and AirPods, each on a stand with Apple's logo, with a presenter in front of a backdrop bearing Apple's bitten apple. The fake iPhone included Apple's native app icons, but Muse hallucinated labels: it named the App Store "Photos" and called Maps, Music and the actual Photos app "Calender." Muse refused to generate an image of an "Apple CEO."
The reporter then opened the "Feed" tab, where Muse generates a news feed based on a text prompt. Using Meta's default prompt, Muse returned AI-generated summaries of current events, including Apple's iPhone Duo launch event, dangerous surf hitting California's coast, and tankers destroyed in the Iran war. The app also listed several news stories from the state where the reporter was visiting family, which Muse admitted it pulled from the shipping address on the Amazon order.
Asked what else it might know based on the Instagram and Facebook accounts used to sign up, Muse listed specific interests including anime, CrossFit, Labrador retrievers, Florida wildlife and '90s/2000s nostalgia, many of the same topics that frequently appear in the reporter's Instagram Reels feed. When asked where it got the information, Muse wrote that Instagram does not show that exact list in the app and that it read it through the account's API data, which "surfaces more than the UI does." It pointed to Instagram's ad topics as the closest visible equivalent. The reporter checked the ad topics list and found vague categories such as workouts, music and electronics, not the same level of detail Muse provided. The Verge reached out to Meta about the data Muse can access inside Instagram and Facebook but did not immediately hear back.
Like other agentic AI tools the reporter has tried, Muse does a decent job performing low-stakes tasks. The reporter also liked that it generates a running list of "ideas" with personalized suggestions.