Meta debuts its Muse AI agent. Will consumers trust it?
Developing story first seen 2 hours ago
Meta has launched Muse, a personal AI agent designed to carry out everyday tasks rather than simply answer questions, marking the company's biggest push yet into consumer AI. The launch comes less than two weeks after Meta agreed to an $18 billion multistate settlement over harms linked to its social media platforms, meaning Muse will need to win a significant degree of user trust, since it requires access to email, calendars, payment details and other personal apps to function.
Muse connects to services such as health and fitness apps, smart home devices, shopping, dining and payments, letting users opt in one service at a time for added transparency; where no direct connector exists, it can use provided credentials or fall back to accessing a service via browser. Powered by Meta's Muse Spark model, it can send emails, book travel, lower bills, fill out forms and make purchases via Link by Stripe, with Shop Pay and 1Password integrations planned. It launches in the US on the web, iOS, Android and WhatsApp, with support for Meta's AI glasses to follow, and is free with a usage meter, backed by two paid tiers, Power at $20/month and Maximum at $100/month, for heavier use; it joins a growing field of consumer AI agents including Google's Gemini Spark and Anthropic's Claude Cowork.
- Meta launches Muse, an AI agent that performs tasks, not just chat
- Requires deep access to email, payments and personal apps
- Free tier plus $20 and $100 monthly paid plans available
New here? Start with this
Meta has rolled out Muse, a new AI system that acts as a personal agent rather than a chatbot, meaning it can carry out tasks such as sending emails, booking travel or making purchases on a user's behalf, rather than just answering questions. To do this, it needs access to sensitive personal information, including email, calendars and payment details.
The launch comes shortly after Meta agreed to pay $18 billion to settle claims across several US states over harms linked to its social media platforms, which is part of why the question of user trust looms large over this release. Meta is one of several major tech companies, including Google and Anthropic, competing to build these so-called "agentic" AI tools that act on people's behalf rather than simply providing information.
Muse is being introduced in the United States on the web, iOS, Android and WhatsApp, with support for Meta's AI glasses expected later. It is free for basic use, with paid subscription tiers available for people who want higher usage limits.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Advocates for agentic AI like Muse argue that consumers stand to gain real convenience and time savings from a system that can book travel, fill in forms and manage purchases on their behalf, and that Meta's phased, opt-in-per-service design is a reasonable, transparent way to let users control what data they share. They would also point out that trust is best earned incrementally through real-world use and demonstrated reliability, and that penalising innovation because of a company's past mistakes risks denying consumers legitimate benefits and ceding ground to rivals who face the same trust hurdles.
The case against
Sceptics argue that handing an AI agent access to email, calendars, health data and payment details represents a significant expansion of a company's reach into consumers' daily lives, and that Meta's recent $18 billion settlement over social media harms is directly relevant evidence of how the company has handled user wellbeing and data in the past. They would contend that opt-in toggles do not fully address deeper concerns about data use, error-prone autonomous actions such as unauthorised purchases, and the difficulty of holding an AI agent accountable when something goes wrong, and that caution rather than early adoption is the more responsible stance until a stronger track record is established.
Full account
Meta has unveiled Muse, a new artificial intelligence agent designed to handle everyday tasks on behalf of consumers, marking the social media giant's most ambitious consumer AI product to date. The launch comes just under a fortnight after Meta agreed to an $18 billion (roughly £14 billion) multistate settlement over the harms caused by its social platforms, a timing that has not gone unnoticed given how much more personal data Muse will require access to compared with a typical social feed. Announced on Tuesday for US users, Muse is pitched as a shift away from the conversational chatbots that have dominated the AI landscape since ChatGPT's rise, towards an assistant that can actually carry out actions rather than simply answer questions.
According to Meta, Muse can be connected to a wide range of everyday apps and services, including email, calendars, payment tools, and platforms covering health, fitness, smart home devices, dining, shopping, music and events. Once given a goal, the agent is said to be capable of working independently, opening a browser, filling in forms, and even negotiating on a user's behalf, continuing tasks in the background after the app has been closed and only returning to the user when approval is needed, such as before completing a purchase. Cited examples of its abilities include drafting and sending emails, booking travel, reducing household bills, converting recipe videos into shopping lists, organising invitations, and making purchases directly through a Stripe-powered checkout, which offers buyer protections intended to reassure users about letting an AI spend money for them. Support for further payment providers is said to be in the pipeline.
Muse is built on Meta's own model, Muse Spark, and will be accessible via a dedicated website, mobile apps for iOS and Android, and through WhatsApp, with support for Meta's smart glasses promised at a later date. Users must link apps and services individually rather than granting blanket access, which Meta frames as a deliberate effort to keep the opt-in process transparent, and a payment card is required to sign up even though basic use is free. Heavier users can pay for one of two subscription tiers offering expanded usage.
The launch places Meta in direct competition with a fast-growing field of task-performing AI agents from rivals including OpenAI, Anthropic, Microsoft, and others racing to move beyond simple chat interfaces, with Google said to be pursuing a comparable consumer-facing assistant of its own. Meta is betting heavily on ease of use as its point of difference, promoting Muse as usable straight away without any technical know-how, and highlighting its ability to remember details from a single mention to inform future suggestions. Underlying all of this, however, is the question of trust: for Muse to be genuinely useful, users must hand over significant amounts of personal and financial information, a big ask so soon after Meta's costly legal reckoning over the way it has handled user data and wellbeing on its other platforms.
Where outlets differ
TechCrunch frames the story primarily through the lens of trust and privacy risk, explicitly linking the launch to Meta's recent $18 billion settlement over social media harms and stressing that Muse demands an unprecedented level of personal data access; it also provides concrete pricing detail (Power at $20/month and Maximum at $100/month) that the other source omits.
The Verge places more emphasis on competitive positioning, situating Muse within a crowded field of rival agents (ChatGPT Work, Claude Cowork, Copilot Tasks, Grok Bot, Moltbot, Gemini Spark) and highlighting Meta's marketing claims about ease of use and lack of a learning curve, while noting Meta did not disclose specific subscription pricing or free-tier limits at the time of its report.
TechCrunch details more of the practical mechanics, such as how Muse connects to services lacking a public API via browser access, and specific integrations like Stripe's Link, Shop Pay and 1Password; The Verge focuses more on Muse's autonomous, background-working behaviour and its memory of user-shared details.
More coverage