Less than two weeks after Meta agreed to a massive deal $18 billion settlement in multiple states over social media consumer harm, company announces biggest bet yet on consumer AI – And it requires much more trust than traditional social media. The company on Tuesday announced Muse, a new personal AI agent that helps consumers with their daily tasks and projects for U.S. users.
To use Muse, consumers must trust Meta with more personal information than ever before. AI agents work by connecting to the users’ apps and services that are part of their daily workflow, such as email, calendar, and payments, as well as apps that individuals may use regularly, such as apps for health and fitness, smart home, dining, shopping, music, events, and more.
The idea is a big bet on what will happen after the ChatGPT era, when AI chatbots answer questions, act as sounding boards, and become digital companions. Instead, Muse focuses on AI that can actually do things for you.
The company says its agents can send emails, book travel, reduce bills, fill out forms, create plans, turn recipe reels into grocery lists, send party invitations, check out with Link by Stripe, and make purchases. The latter offers purchase protection, which could ease consumers’ concerns about having an AI checkout on their behalf. (Shopify’s Shop Pay integration with 1Password is also coming soon.)
Muse users can decide which apps and services they want to connect to, one at a time, making the opt-in nature of their use of Muse more transparent. The agent is powered by Muse Spark, Meta’s AI model, and comes with built-in connectors (pictured below) for several services, with more to come. If the service you want is not available, but a public API is provided, Muse can set up a connection using the credentials you provide. If no API is available, Muse can access the service via the browser instead.
Muse will initially be available through the web (muse.ai), iOS and Android apps, and WhatsApp chats. The company says it will soon be featured in Meta’s AI glasses as well. You’ll need a payment card to start using Muse, as it’s free to use and a subscription plan kicks in as your usage increases.
Two paid plans will be available at launch. Power costs $20 per month and Maximum costs $100 per month. Both of these subscriptions give you more Muse usage to take over your day-to-day tasks, but Meta believes most people will remain on the free tier. (According to the company, the app includes a usage meter that shows users what percentage of their usage is left. It also alerts users when they run out of free usage and provides subscription options.)
Like other AI agents, Muse continues to work even after the user leaves the app. It will also improve over time by learning from user conversations what is important to make unprompted suggestions, Mehta noted.
This concept is not unique to meta. The agent age is here in earnest, with large companies and small startups alike experimenting with how AI agents make the most sense for consumers and how they can be integrated into people’s daily lives.
Some have tried AI web browsers and services that can initiate various tasks on behalf of consumers, such as Gemini Spark and Claude Cowork. Some companies are integrating AI into the chat applications that consumers use most, such as Apple’s iMessage platform, SMS, and WhatsApp.
Despite their usefulness, these powerful agents are forcing consumers to wrestle with difficult questions about how much privacy they are willing to give up. Many early testers of AI assistant Instinct were shocked to find that the app required a broad “perpetual and irrevocable” license to “access, use, host, cache, store, reproduce, transmit, display, publish, distribute, and modify” user materials, including training AI models.
Meta claims that internally, Muse runs on its Muse Secure VM, a “dedicated secure computer with its own browser,” which provides a range of privacy, safety and security protections for customer data. The company says a separate Sentinel agent runs on the same virtual machine but is separate from Muse at the system level.
This means Muse has no way of knowing your passwords or payment methods. Meta also claims that Muse does not share people’s conversations or data with Meta’s advertising system.
(These claims are detailed in a technical post also released today, but require deeper investigation by security experts.)
Could Meta’s history negatively impact Muse’s adoption?
Despite Meta’s documented security measures, it remains to be seen whether the company has enough consumer trust to make its agents successful.
As it stands, the meta has a history of declaring one thing and doing another. For example, in 2011, the tech giant settled with the FTC for deceiving consumers by disclosing their personal information without their consent. In 2019, the FTC fined Facebook for eight separate privacy violations in a then-record $5 billion settlement. In 2023, the FTC charged Mehta with violating a privacy order filed after a 2019 settlement.
When it comes to technical issues, Meta has had major failures in the past, with a large number of users’ passwords being discovered in readable format in 2019, leaving people exposed to potential hacks. The massive Cambridge Analytica data scandal, in which Facebook data belonging to millions of consumers was collected by third parties without their consent, is still fresh in some people’s minds.
Mehta has also been repeatedly brought before Congress to testify about how he protected, or failed to protect, minors from harm. Because Congress failed to act, meth ended up becoming the target of several related lawsuits. These include a lawsuit Mehta just settled with 29 states in August, a New Mexico child harm lawsuit in which Mehta was ordered to pay $942 million, and thousands of personal injury and school district lawsuits pending against several social media giants.
To allay consumer concerns, Meta not only details its security promises, but also provides technical documentation and explanations. The company also designed Muse in a way that allows consumers to feel more connected to the agent itself. Users can customize their Muse by giving it a name, choosing an avatar, and configuring various settings that determine its appearance and how it communicates with the agent.
Time will tell whether this personal connection and the utility that Muse provides is enough for consumers to trust Meta again with their personal information.
If you make a purchase through links in our articles, we may earn a small commission. This does not affect editorial independence.
