The company just settled a massive child-safety lawsuit. Now it's asking users to hand over their most personal data

If timing is everything, Meta may have picked one of the worst possible moments. Just days after agreeing to an $18 billion settlement with 29 U.S. states over allegations that Facebook and Instagram harmed children, the company has rolled out its most ambitious consumer AI product yet — an AI agent that wants access to your email, calendar, payment details, and even your home security cameras.

What Exactly Is Muse?

On Tuesday, Meta unveiled its new personal AI agent app, called Muse — internally codenamed "Hatch." It's built on Meta's own Muse Spark family of foundation models, which AI chief Alexandr Wang has been rolling out at a rapid pace since April.

Zuckerberg introduced it himself on X, describing it as a personal agent that understands your goals and works around the clock to get things done. In practical terms, that means Muse can send emails, book travel, try to lower your bills, fill out forms, turn recipe videos into grocery lists, and even send out party invitations.

How Does Meta Plan to Make Money From It?

Muse will come in three tiers — a free option, plus two paid plans: "Power" at $20 a month and "Maximum" at $100 a month, depending on usage. The company expects most users will stick with the free tier.

Meta is also exploring ways to earn from shopping transactions completed through Muse, meaning it could eventually take a cut whenever the agent makes a purchase on someone's behalf — though no concrete plan has been finalized yet.

The Big Question: Trust

This is where the real challenge begins. For Muse to work, it needs access to your email, calendar, payment services, smart-home devices, and health apps — essentially most of your digital life. Meta says users will connect each service individually, one at a time, to keep the process transparent.

Vishal Shah, the company's VP of AI products, claims the engineering team designed the product to meet strict standards for privacy, security, and performance — though he acknowledged that no system can guarantee it will never make a mistake.

Each instance of Muse runs inside its own isolated virtual machine. Meta says the agent never sees actual passwords or payment details, and asks for permission before doing anything sensitive. A separate "Sentinel" agent also runs on the same virtual machine but is kept fully separate from Muse at the system level.

Could Meta's Track Record Hurt Muse's Chances?

For Meta, the real test isn't technology — it's trust. The company's privacy history has been rocky for over a decade: an FTC settlement in 2011, a record-breaking $5 billion fine in 2019, and further FTC charges in 2023 over violating an earlier privacy order. Add to that the lingering memory of the Cambridge Analytica scandal, and skepticism seems inevitable.

On child safety specifically, Meta has repeatedly testified before Congress, and a New Mexico court recently ordered the company to pay over $942 million in a related case.

Given all that, the obvious question is: will people really be willing to hand over nearly every corner of their digital life — minus passwords and payment info — to a company with this kind of track record?

Part of a Bigger Industry Shift

Meta isn't alone in this push. Google's Gemini Spark and Anthropic's Claude Cowork are already experimenting with similar agentic capabilities. Notably, Muse launched in the same week Apple is expected to unveil a more capable, chatbot-style version of Siri.

The broader tech industry is moving beyond chatbots that simply answer questions, toward agents that actually act on your behalf. But as these agents gain more power and access, the privacy and security stakes grow just as fast.

For now, Muse is only available in the U.S., and works across iOS, Android, the web, and WhatsApp, with support for Meta's Ray-Ban smart glasses coming later. The real question that remains: will users give Meta another chance?