Less than two weeks after Meta reached a massive $18 billion multistate settlement over claims about social media’s harm to consumers, the company unveiled its largest bet yet on consumer AI—one that demands far more trust than social media ever required. On Tuesday, Meta introduced Muse, its new personal AI agent designed to assist U.S. users with everyday tasks and projects. To use Muse, consumers will need to hand over more personal information to Meta than ever before. The agent works by linking to the apps and services people already use in their daily routines—email, calendars, payments, and other frequently used tools like health and fitness apps, smart home devices, dining, shopping, music, events, and more. This marks a significant gamble on what comes after the ChatGPT era, where chatbots answered questions, acted as sounding boards, or became digital companions. Instead, Muse focuses on AI that can actually take action for you.

According to the company, the agent can handle tasks such as sending emails, booking travel, negotiating bills, filling out forms, creating plans, turning recipe reels into grocery lists, sending party invitations, and making purchases—using Link by Stripe for checkout. That payment option includes purchase protections, which might help ease consumer worries about letting an AI handle transactions on their behalf. (Shopify’s Shop Pay and 1Password integrations are also expected soon.)

Muse users can choose which apps and services to connect, doing so one at a time, to make the opt-in nature of the experience more transparent. The agent runs on Meta’s AI model, Muse Spark, and comes with built-in connectors (shown below) for several services, with more planned over time. If a desired service isn’t available but offers a public API, Muse can establish a connection using credentials provided by the user. When no API exists, Muse can access the service through the browser instead.

Initially, Muse will be available on the web at muse.ai, through iOS and Android apps, and via chats in WhatsApp. It will soon expand to Meta’s AI glasses, the company says. The service will be free to start, with subscription plans introduced as usage grows—which is why Muse requires a payment card upfront. Two paid tiers will be offered at launch: Power at $20 per month and Maximum at $100 per month. Both subscriptions provide more Muse usage for offloading everyday tasks, though Meta believes most users will stay on the free tier. (The app includes a usage meter that shows users how much of their allowance remains and warns them before usage triggers payment.)

Like other AI agents, Muse keeps working even after the user leaves the app. It also improves over time by learning from conversations what matters most to the user, allowing it to make unsolicited suggestions, Meta noted.

The concept isn’t unique to Meta. The agentic era is coming into its own, as both large corporations and nimble startups experiment with how AI agents can best serve consumers and weave into daily life. Some have tried AI-powered web browsers or services, like Gemini Spark or Claude Cowork, which can initiate various tasks on behalf of users. Others are embedding AI into the messaging platforms people already rely on, such as Apple’s iMessage, SMS, and WhatsApp.

Despite their usefulness, these powerful agents have forced consumers to confront tough questions about how much privacy they’re willing to sacrifice. Many early testers of the AI assistant Instinct were alarmed to discover the app demanded a broad “perpetual and irrevocable” license to “access, use, host, cache, store, reproduce, transmit, display, publish, distribute, and modify” any user materials, including for training its AI models.

Under the hood, Meta says Muse operates in its own “dedicated, secure computer with its own browser,” called Muse Secure VM, which provides various privacy, safety, and security safeguards for customer data. The company notes that a separate Sentinel agent runs on the same virtual machine but is kept isolated from Muse at the system level.

This design means Muse won’t have access to users’ passwords or payment details. Meta also claims that Muse doesn’t share conversations or data with Meta’s advertising systems. (These assertions are laid out in more detail in a technical post released today, though they’ll require deeper scrutiny from security experts.)

Could Meta’s track record undermine Muse adoption? Despite Meta’s documentation of its security measures, it’s unclear whether the company has enough consumer trust for its agent to succeed. Historically, Meta has a pattern of promising one thing while doing another. In 2011, for example, the tech giant settled with the FTC over charges that it misled consumers by making private information public without consent. In 2019, the FTC fined Facebook a then-record $5 billion for eight separate privacy violations. In 2023, the FTC accused Meta of breaching a privacy order filed after the 2019 settlement. On the technical side, Meta has also stumbled—such as in 2019, when it discovered that a number of users’ passwords were stored in readable formats, leaving people vulnerable to hacks. The massive Cambridge Analytica scandal, where data from millions of Facebook users was harvested by a third party without consent, still lingers in many people’s minds as well.

Meta has also been repeatedly called before Congress to testify about how it protected—or failed to protect—minors from harm. With lawmakers failing to act, Meta became the target of several related lawsuits, including the one it just settled with 29 states in August, a New Mexico case over child safety that resulted in a $942 million judgment, and the thousands of personal-injury and school-district lawsuits still pending against multiple social media companies. To ease consumer concerns, Meta not only talks at length about its security commitments, offering technical documentation and explanations, but has also designed Muse to foster a sense of personal connection. Users can customize Muse by naming it, choosing an avatar, and configuring its appearance and settings that govern how the agent communicates. Only time will tell whether this personal touch and the practical value Muse offers are enough for consumers to once again trust Meta with their personal information.