Meta's Muse AI Agent Surpasses ChatGPT Launch Metrics Amid E-Commerce Clashes
Meta's new autonomous AI agent app, Muse, has surpassed ChatGPT's early iOS download metrics, sparking industry excitement and e-commerce friction with Amazon while raising critical cybersecurity concerns.
Autonomous AI agents are rapidly becoming the new frontier of the technology world, and Meta appears to have found its next major hit. Muse, the tech giant's personal agent app launched last week, is exhibiting extraordinary growth figures.
According to preliminary data from market intelligence firm Apptopia, during its first 12 days on iOS in the US and Canada, Muse recorded approximately 1.8 million downloads, compared to just 1.3 million for ChatGPT during the exact same timeframe following its iOS launch. In total, approximately 2.8 million global installations were recorded across all app stores, with research firm Appfigures also noting an early crossing of the one million download milestone.
E-Commerce Friction: Amazon Closes the Door, Shopify Opens It
A core capability offered by Muse is agentic shopping—the ability to execute online purchases on behalf of the user. However, attempting to operate on e-commerce sites encountered immediate resistance from the industry's largest player. Muse users attempting to make purchases on Amazon encountered an unusual error message stating: "Ongoing access by an unauthorized AI agent violates Amazon's terms of service, which our customers agreed to."
This blocking is explained not only by the technological competition between the two cloud and AI giants, but also by Amazon's practical concerns: in the event that an AI agent executes an incorrect order due to a hallucination, Amazon would be forced to deal with the fallout involving angry customers and suppliers.
Practical Applications and Expert Workflows
Muse extends beyond traditional chat interfaces, focusing instead on executing practical actions in the background. Content creator Peter Yang described it as the best personal agent he has tested so far, noting that the app saved him over 800 dollars annually.
It saved me over 800 dollars on cable and phone bills. The core capability in my view is customer service handling—Muse is capable of navigating phone voice menus, waiting on the line, and conversing with a human support representative while representing you as an agent.
Security Risks of Autonomous Agents
Alongside the excitement, the transition from passive language models to autonomous agents operating in the real world raises severe security questions. For an agent like Muse to execute tasks such as canceling subscriptions, booking flights, or sorting an inbox, users are required to grant it broad access permissions—including connection to primary email accounts, bank accounts, and occasionally credit card details.
As explained by Eli Smadja, a director at Check Point's research department: "You need to look at an AI assistant almost like another user inside our account. If it has access to email, files, financial information, or other applications, any attacker who successfully manipulates the agent will find themselves with a 'master key' in hand that enables access to all these resources. The principle of least privilege becomes critical here: giving the assistant only the access it truly needs, and nothing more."
