Meta (META.US) officially launched its AI personal assistant app, Muse, on Tuesday, marking its first attempt to charge subscription fees to some users. As the AI industry faces mounting pressure over safety, privacy, and the returns on massive capital expenditures, the tech giant is striving to evolve AI from a simple chat tool into a "personal agent" capable of executing tasks directly for users, positioning it as a cornerstone product and revenue source for the next phase of growth.
Developed under the internal codename "Hatch," Muse is built on the Muse Spark series of foundation models, introduced since April by Meta's AI chief, Alexandr Wang. Unlike traditional chatbots that primarily offer information and answers, Muse focuses on completing real-world digital tasks for users, including booking services, filling out spreadsheets, and even monitoring home security cameras. Wang explained that Meta aims to make Muse feel simple, friendly, and easy to understand for the average user. While users see a straightforward interface, Muse may be executing complex coding tasks, building integrations between various services, and handling numerous sophisticated operations behind the scenes.
Free Tier and Two Paid Plans Offer Options Up to $100 Per Month
Meta plans to offer Muse with a free usage tier alongside subscription plans priced at $20 and $100 per month, depending on user consumption levels. This pricing model signals Meta's initial attempt to generate AI revenue directly from consumers. Mark Zuckerberg has repeatedly signaled to investors that more powerful personal AI assistants will become the company's next major wager. These agents are envisioned to not only help users organize emails but also proactively seek out product deals and manage routine daily affairs. During Meta's earnings call in July, Zuckerberg said new personal agents would form the "basis for the next wave of products and revenue sources" in the coming months and years. This strategy explains Meta's sustained heavy investment in data centers, computing power, and other AI infrastructure. As capital expenditures continue to climb, Wall Street is increasingly eager to see these investments translate into tangible revenue, not just more capable AI models.
Muse is not Meta's only recent AI commercialization attempt. The company has previously launched the Muse Code AI agent for developers and is testing multiple AI subscription services to establish new revenue streams beyond its traditional digital advertising business.
Beyond Chat: Executing Tasks Like Booking, Form-Filling, and Shopping
The key differentiator for Muse lies in its "action capability." Users can authorize the agent to perform a range of real-world tasks, rather than merely asking it for guidance. For example, Muse can assist with booking appointments, filling out online forms, and executing operations across different digital services. The Muse app also includes a feed that aggregates updates from users' Facebook and Instagram accounts, along with articles and information from across the internet, which the AI then organizes and summarizes. Muse will also be available on WhatsApp, though this version will primarily retain core conversation and agent functionalities, omitting the full feed found in the standalone app and the "Ideas" tool that recommends AI agent use cases.
Meta stated that Muse will first launch in the United States, accessible via iOS, Android, and a standalone website, with future plans to integrate it into Ray-Ban Meta smart glasses.
Meta Explores AI Shopping Commission, Personal Agents May Become a New Commercial Gateway
Beyond subscription fees, Meta is researching a potentially more significant business model: taking commissions from shopping transactions facilitated by its AI agents. Wang noted that because personal agents can help consumers find products of interest and assist with completing purchases, the commercial transaction model holds considerable potential for Muse. However, Meta has yet to finalize specific fee or revenue-sharing mechanisms. If this model comes to fruition, Muse's significance to Meta could extend far beyond the $20 or $100 monthly subscription revenue. Traditional internet platforms primarily direct consumers to merchants through search, social media feeds, and advertising, whereas AI agents may directly participate in the entire "find product, compare prices, make choice, complete transaction" process. This implies that platforms controlling the personal agent gateway could further insert themselves into the internet's commercial transaction chain. For Meta, which still relies heavily on digital advertising for revenue, this represents a potential new growth direction.
AI Agent Security Risks Under Scrutiny; User Data Defaults for Model Training
Nevertheless, as Muse launches, the safety and privacy concerns surrounding AI agents are facing increasingly stringent scrutiny. Compared to standard chatbots, personal agents require elevated permissions to access user accounts, information, and third-party services in order to perform actions on users' behalf. Should the model be compromised or execute instructions erroneously, the potential consequences are more severe. Wang stated that Muse operates within an isolated environment on Meta's computing infrastructure, never sees users' actual passwords or payment information, and requests user permission before performing sensitive operations.
Notably, Meta utilizes an opt-out mechanism for Muse user data, meaning users must actively choose to opt out to prevent Meta from using their interactions with Muse to train its AI models. Meta's engineering vice president, David Singleton, said that if users do not opt out, the company will remove "key personally identifiable information" from the data before using conversation and interaction data to improve its models. Meta is also inviting third-party security researchers to identify vulnerabilities in Muse through a bug bounty program, offering rewards for qualifying findings. The company said it has strengthened Muse through extensive internal testing, agent red-teaming, and addressing security issues discovered in real adversarial environments prior to the official launch.
Tech Giants Vie for the 'Personal Agent' Gateway
Meta's release of Muse also reflects a shift in competitive dynamics within the AI industry. With the rapid rise of OpenAI's Operator (referenced as OpenClaw in prior text, implying a general trend), AI agents capable of autonomously invoking models, tools, and software to accomplish complex tasks have become a new focal point of competition in the tech sector. OpenAI, Google, and startups like Town are all releasing products or features in this domain. Wang said that compared to some competing products, Muse aims to lower the barrier to entry for personal agents, making them accessible to a broader mainstream consumer market. However, he also acknowledged that the industry is still in the very early stages of the personal agent era.
Meta's launch of Muse also comes with added pressure. The company recently agreed to pay nearly $17 billion to settle litigation with a coalition of state attorneys general over potential harms related to Facebook and Instagram platforms, while still facing personal injury and school district-related lawsuits. Concurrently, skepticism in American society is rising over AI data center construction, energy consumption, and potential AI-related cybersecurity risks. Muse represents not just a new AI product for Meta but a significant test of whether its massive AI spending can generate real commercial returns. If Meta can accustom everyday consumers to entrusting booking, shopping, information organization, and even household digital tasks to AI, the personal agent could become another critical internet gateway for Meta, following its social media feed. Subscription fees, AI shopping commissions, and future integration with smart glasses may collectively constitute the core path of Meta's next-stage AI commercialization.
Comments