Meta CEO Mark Zuckerberg and Meta President and Vice Chairman Dina Powell McCormick attend the annual Allen & Company Sun Valley Media & Technology Conference held at Sun Valley Resort in Sun Valley, Idaho, USA on July 9, 2026.
Brendan McDiarmid | Reuters
Meta introduced its AI personal agent app on Tuesday and is asking some users to pay a fee as the company and the broader artificial intelligence industry face intense scrutiny and public skepticism.
The app, which has been internally codenamed “Hatch,” leverages the Muse Spark family of foundational models that Meta AI head Alexandr Wang has been debuting at a breakneck pace since April.
Mehta said the Muse app lets you delegate digital tasks to AI-powered assistants, such as booking appointments, filling out electronic forms, and even monitoring home security camera footage. The company designed the app to be “very approachable and friendly and easy to explain and not feel too complicated,” Wang said in an interview with CNBC.
“Behind the scenes, Muse can perform some very sophisticated coding workflows, build sophisticated integrations, and do quite a bit of heavy lifting while keeping things very easy for users,” Wang said.
Wang said Muse Personal Agent is available through a free tier or a $20 or $100 monthly subscription plan, depending on usage.
Meta CEO Mark Zuckerberg, who hired Wang last year as part of a $14 billion investment in his startup Scale AI, has hinted to investors that the company’s next big AI bet will include a powerful digital assistant that scrutinizes emails and looks for deals on furniture.
Zuckerberg has said supercharged digital assistants are the next big leap forward in AI models, touting them as one of the reasons Meta is spending so much on data centers and infrastructure. “The new individual agents will be the foundation of our next product suite and revenue line in the coming months and years,” Zuckerberg said during an earnings call in July.
The launch of Meta’s personal assistant comes at a particularly difficult time for the company. Meta recently agreed to pay nearly $17 billion in a major settlement with a coalition of state attorneys general that had sued the company for misrepresenting the prevalence of harm on apps like Facebook and Instagram. Mehta still faces a wave of personal injury lawsuits and school district lawsuits involving similar allegations.
Meanwhile, the country faces a growing backlash against AI data centers and the companies building them, with growing concerns about the potential cybersecurity risks posed by AI agents and the underlying infrastructure models.
But Meta is also under pressure from Wall Street to show a return on its AI investments as it seeks to diversify its business and reduce its reliance on digital advertising. This personal agent follows other new initiatives, including the Muse Code AI agent for developers and various subscription plans that Meta recently rolled out to create new lines of business.
Addressing security concerns, Wang said the app runs within its “unique isolated environment” within Meta’s computing infrastructure and “never sees or asks for your actual passwords or payment details before doing anything sensitive.”
Users must opt out of Meta using their interactions with Muse personal agents to train the company’s AI models. If it doesn’t, the company will scrub “significant personally identifying information” from agent conversations and interactions before using them to improve its AI models, said David Singleton, Meta’s vice president of engineering.
Wang said the company is exploring possible financial initiatives, such as reducing some shopping transactions related to AI agents, but no concrete plans have been finalized yet.
“We think the commerce business model could potentially be very interesting for this product because it allows people to find what they’re actually excited about and end up making more purchases,” Wang said.
Users of the Muse Personal Agent app can view a feed showing various updates from their connected Facebook and Instagram accounts, or specific articles and stories pulled from around the web, depending on what they want their feed to aggregate and summarize. The agent will also be available via WhatsApp, but that version will not include app features such as a feed or an idea tool that recommends trips and tricks for the AI agent, Wang said.
“The core functionality for key conversations with agents is the same between these two services,” Wang says.
Meta is influencing major trends in the industry. Since the meteoric rise of OpenClaw, which is used by programmers to manage the AI models that power digital assistants, companies like OpenAI and Google, along with startups like Town, have introduced similar agent tools and capabilities.
Wang said Muse Personal Agent was designed to be “more accessible to a wider audience” than competing products, but acknowledged it was still “in the fairly early stages of this new era of personal agents.”
Meta asks third-party security researchers to find vulnerabilities within the Muse personal agent through a so-called bug bounty program that includes monetary rewards for discovering specific issues.
“We enhanced Muse based on issues found in real-world adversarial scenarios by security researchers in extensive dogfooding, agent red-teaming, and a private bug bounty program,” the company said in a technical blog post.
The personal agent will be available to U.S. consumers on iOS, Android or through a standalone website, Meta said. It will eventually be accessible through the company’s Ray-Ban Meta glasses.
WATCH: Colorado says the most important outcome of meth lawsuits is behavior change, not money.
