Meta has launched Muse, an AI assistant built to handle daily tasks and manage personal information. The company says Muse can take “the busywork off your plate” by booking tickets, making appointments, filling out forms, and connecting to email, calendar, or health information when users give commands.
That convenience comes with a major tradeoff: users have to hand over a large amount of trust. Meta has 3.6 billion daily users, and its new assistant is designed to work across information that can reveal where people go, what they buy, who they contact, and what health details they choose to manage through the service.
Meta announced Muse on September 9, 2026, at 1:10pm EDT, and made security a central part of the announcement. CEO Mark Zuckerberg posted a video promoting trust in Muse, highlighting its security, privacy, and trustworthiness. He had posted about trusting Muse on September 8, 2026, at 8:02 PM UTC.
Muse promises control, but raises privacy questions
Meta claims Muse runs on a secure “dedicated virtual machine.” The company also says it will store login credentials in a way that the AI agent cannot access. Those safeguards are meant to separate the assistant from the passwords and account details it needs to help users complete tasks.
Still, testing showed how much information Muse could uncover. Emma Roth found an “unnerving amount of information” about herself, including data from her Gmail and Amazon accounts. Muse figured out the state where Roth was visiting her family and pulled information from her Amazon order shipping address.
The assistant also listed Roth’s “very specific interests” at a level beyond typical ad targeting. That finding turns the privacy question into something more personal than password protection: users must also consider how much the assistant can connect, infer, and present back from accounts they already use.
Muse is not limited to one app or one type of task. Meta built its AI assistant into Facebook, Instagram, WhatsApp, and Messenger, while Muse itself can connect to email, calendar, or health information when users give it commands. The design puts a broad range of personal details within reach of AI tools.
Invasive prompts add to Meta’s AI problems
Concerns about Meta’s AI features grew after Meta AI prompted Instagram user Kalie Robins with invasive questions about her children. The prompts appeared after Robins posted a video of herself and her child singing in the car.
Meta AI suggested questions about Robins’ children’s ages and location. It also showed photos of her daughters posted on Instagram, including one Robins said she had deleted years ago.
Meta spokesperson Dina El-Kassaby said, “The feature never should have prompted the individual with questions like that.” She also said, “We created this feature to help people get more information about topics or posts that interest them,” based on information the user already had access to.
Meta pulled a feature in July that allowed people to create AI deepfakes of other Instagram users after facing backlash. El-Kassaby said the company “missed the mark,” underscoring the gap between Meta’s intended use for AI features and the way those features can affect people in practice.
Trust is now part of the product
Muse’s launch makes trust part of the product itself. Meta is asking users to let an AI assistant work across accounts, complete tasks, and handle information tied to communication, shopping, scheduling, and health. The company says its technical protections keep login credentials away from the AI agent, but the assistant’s reach still depends on the access users grant it.
The timing also puts Muse alongside fresh questions about Meta AI’s prompts and data use. On September 11, 2026, Meta published an article at 1:09 PM EDT, while a report at 2:25 PM UTC addressed Meta changing AI suggestions. The company’s response to Robins’ experience shows that Meta is adjusting how these suggestions work after an AI feature crossed a line.
That line matters because Muse is built to act, not just answer questions. Booking a ticket, making an appointment, or filling out a form requires access to personal details and accounts. When the same system can also identify a family’s location, pull an old photo, or describe highly specific interests, users are left judging whether the convenience is worth the trust.
Meta wants Muse to become a helpful layer across its services and personal accounts. Its security promises and Zuckerberg’s message ask people to trust the system. The experiences involving Roth and Robins show why that trust will depend on more than secure storage: users will also watch what Muse can discover, what it chooses to ask, and how much of their private lives it brings into view.
Based on
- Meta Releases Uber-Creepy AI Chatbot as Its Platforms Crumble Under Grotesque Child Abuse — futurism.com
- Meta says it’s changing AI suggestions after posing invasive personal questions | The Verge — theverge.com
- Meta launches Muse AI assistant | Semafor — semafor.com
- Muse: Meta launches new AI agent – here’s what to know | The Independent — independent.co.uk
- Mark Zuckerberg tells you to trust Muse. | The Verge — theverge.com
