Muse is new AI sensation, it is helpful, from Mark Zuckerberg and potentially dangerous

Mark Zuckerberg's new AI agent, Muse, is making waves online because of how it can help you run your digital life. But the Muse AI agent has also raised privacy concerns from some over user data. Here is everything you need to know about Meta Muse.Meta CEO Mark Zuckerberg says Muse is your personal AI agent that can work around the clock. (Photos: Reuters/Meta)Earlier this month, Meta released Muse – a new AI agent that is designed to be your personal assistant. At the time of its release, Meta CEO Mark Zuckerberg wrote that Muse “understands your goals and works 24/7 to get things done for you.” Now, just about two weeks later, Muse has become one of the most popular AI tools, reaching the No. 1 free app spot on the App Store in the US.Meta has pitched Muse as a personal agent that can keep working in the background after you give it instructions, unlike a conventional chatbot.What is Meta Muse?The Muse personal agent can do a lot more than a chatbot like ChatGPT. It can send emails, book flight tickets, manage schedules, and even shop online for you. To do all of this, though, Muse does need to connect with services such as email, calendars, payment platforms, shopping apps and smart-home systems. Muse runs on Meta’s Muse Spark AI model that was developed by the company’s Superintelligence Labs, led by Alexandr Wang.Meta is offering the personal agent in a free tier and paid plans priced at $20 (roughly Rs 1,915) and $100 (roughly Rs 9,579) a month. In case you are wondering whether Meta has simply copied OpenClaw – the open-source agentic AI platform that is widely used today – OpenClaw's creator Peter Steinberger has clarified that it is not the case. Steinberger wrote on X, "Because this “Meta uses OpenClaw” story is going around. They built their own agent, being inspired. Nat and his team did a great job. Kudos!" OpenClaw creator Peter Steinberger cleared the air over rumours of Meta copying OpenClaw. Alexandr Wang also shared a screenshot on X from September 2025, showing that his team had been planning to build a personal AI agent since then. "If you wanna know how long we've been working on muse, here's an excerpt from a doc that I wrote for the meta board a year ago in sep 2025," he wrote. "The whole team & i have been building muse for a long ass time." Alexandr Wang shared on X. Internet is loving Meta MuseUnlike most coding-centric AI agents that we have seen from OpenAI and Anthropic, Meta’s Muse is more focused on your personal activities. This could make Muse more appealing for those who simply want an AI assistant to make their life more convenient – by letting the AI agent shop for them or book tickets for a trip. Though OpenAI and Anthropic also offer similar capabilities.Setting up Muse is also fairly easy. The Meta AI agent is available to adults in the US through a standalone app and on WhatsApp. In the app, you can pick an avatar for your AI agent and then connect it with the services you want to use. You can also transfer info from Facebook, Instagram, or Threads, helping the AI learn more about you. Once done, you can send prompts to Muse and the AI does the rest.As per Sensor Tower, Muse logged 730,000 downloads in about five days after its September 8 launch. Muse's cumulative downloads in its first 13 days were above Claude's 400,000 and Grok's 200,000, though still behind ChatGPT's 3.1 million during a comparable period.Following Muse’s launch, Meta shares rose more than 11 per cent on Monday as investors gained confidence over the company’s AI spending. Investment firm Evercore ISI said, "Meta has a hit on its hands with Muse.” Mark Zuckerberg’s net worth reportedly increased by $25 million, going up to $253.6 billion.But there are concernsAt the same time, Muse’s rollout has added to wider concerns around AI safety, cyber security and privacy. As we mentioned previously, Muse needs access to your personal data such as emails and other services to get tasks done, raising privacy fears. Though Mark Zuckerberg insists, “Muse is built from the ground up for privacy and security. Your data and credentials live on the Muse Secure VM -- an isolated Linux computer with a browser, CPU, memory, and storage.”Meta states that users can choose which apps Muse can access, and revoke permissions at any time. You can also stop conversations from being used to train AI models.As per a report by Wired, during one test, the Muse AI agent repeatedly suggested adding more personal data or including access that would let it read financial information. The AI agent also suggested monitoring inboxes or using meal photos to estimate calories.While Meta states that any user data used for training will be “sanitised” to remove identifying information, there are concerns that context pulled from connected sources such as email inboxes or bank accounts could still be involved. The company has not clarified how the sanitisation process works.Fresh concerns have also come from a reported zero-day vulnerability in Muse's macOS app. Ars Technica reported that macOS security researcher Patrick Wardle found that any local app or terminal command could change undocumented Muse settings, including the endpoint used for transcription. By redirecting that endpoint to a server they control, a hacker could capture the authentication token for a user's Muse account and gain complete control of the agent."We can manipulate the agent and leverage its privileges to do whatever we want," Wardle told Ars. He said proof-of-concept attacks could write malicious files to disk and take pictures, in many cases without any warning to the user. Wardle said Meta's decision to send dictation to the cloud, instead of using on-device macOS tools, and its decision to let any app alter undocumented settings made the exploit possible.The growing use of such agents is already affecting online platforms. Amazon has blocked Muse from its shopping site, citing privacy and security risks. The company said it had not been given advance notice, and alleged that Muse bypasses personalisation features, appears to capture and store customer credentials, and scrapes account data. In a statement, an Amazon spokesperson said, "We think it's fairly straightforward that third-party applications that offer to make purchases on behalf of customers from other businesses should operate openly and respect service provider decisions about whether or not to participate."Shopify, however, is working with Meta, with chief executive Tobias Lutke announcing a partnership to allow agentic checkout in Shopify stores.- EndsPublished By: Armaan AgarwalPublished On: Sep 22, 2026 08:57 IST