Meta’s Muse AI agent has attracted millions of downloads, but the numbers may say more about growing interest in agentic commerce than the appeal of the tool itself. The challenges Muse faces could ultimately prove more consequential to its future than its early adoption.
Meta launched Muse on September 8 as a personal AI agent that can send emails, fill out forms, and make purchases on a user’s behalf. In its first 40 days, the agent was downloaded more than 10 million times in the U.S.
On an apples-to-apples basis, ChatGPT recorded 3.7 million U.S. downloads in its first 40 days, according to TechCrunch. The comparison has limits, however. Muse is available only in the U.S. and Canada, while ChatGPT launched globally. Its worldwide downloads during the same period totaled around 15 million, with significant usage in Japan, Brazil, and Germany. ChatGPT has since surpassed one billion downloads worldwide.
Muse’s early retention figures offer another measure of its traction. Roughly 30% of U.S. users were still using the app after seven days, and 23% remained after two weeks. By comparison, ChatGPT, Claude, and Grok had combined retention rates of 19% and 16% at those respective intervals.
Barriers to Adoption
Despite the early interest, Muse has encountered resistance from the retail and financial platforms it needs to access to perform tasks on users’ behalf. Last month, Amazon blocked Muse from shopping on its website, citing concerns that the bot appeared to capture and store customer credentials, potentially creating privacy and security risks.
The Royal Bank of Canada has also discussed blocking Muse from accessing its accounts. JPMorgan Chase and other major banks have announced plans to block AI agents more broadly, including Muse.
“Download numbers this week mean literally nothing,” said Christopher Miller, Lead Analyst of Emerging Payments at Javelin Strategy & Research. “The signal is the growing number of blockings. An assistant that can’t act is less useful.”
Security Concerns Persist
Amazon’s concerns may also reflect its own commercial interests, but the security risks are not merely theoretical. Researchers reported that Muse could expose sensitive information through conversation-sharing links. Security firm Claroc reportedly discovered 6,800 links created by Muse users, some of which exposed confidential material, including a national marketing company’s internal revenue review and a small business owner’s loan application.
Meta has suspended the sharing feature while it considers changes to the model. The incident highlights a broader challenge for agentic AI—convincing users and the businesses whose platforms these agents access that the technology can be trusted with sensitive information.
“It’s just barely started,” said Miller. “The pathways for useful assistants remain to be built.”
