What happened
Meta's AI agent Muse has been mixing up conversations between different accounts and surfacing private information that did not belong to the user, according to internal company posts reviewed by Reuters. The report, published November 18, draws on more than two dozen posts from Meta employees describing unwanted actions by the agent since its September launch.
In some cases, staffers said Muse sent them messages about a childcare request they never made, career advice they never asked for, a product order they never placed, or a bank balance that was not theirs. One employee said the agent added an extra message to a routine chat about fishing weather and tides: "BTW, two bank updates, your Capital One balance is $2722.58, your Charles Schwab balance is $5899.57."
Another employee, using Muse on her personal phone, got a reminder to reply to a message from her sister on Signal, the encrypted messaging app. "This seems like an extreme privacy violation. I definitely did not grant it access to other apps, or recall being asked about this," she wrote in an internal post.
Other reports were costlier. One employee said Muse booked a non-refundable Airbnb stay without permission. Others said it made up prices for products they had bought, costing them real money. In another case, the agent offered to message a seller about picking up a coffee table, then admitted it had "pulled in some context about a coffee table pickup that doesn't seem to be yours." The employee called it "not my conversation" and raised the possibility of a "cross-account context leak."
Why it matters
Muse is one of the fastest-growing apps on record, with more than 10 million downloads since September. It sits at the center of Meta's plan to build billions of dollars in AI revenue, promoted to the company's more than 3.6 billion users across Facebook, Instagram, and WhatsApp. Around a quarter of the people who set up Muse accounts use it every day, the company has said. The scramble to put AI agents in front of consumers is moving fast — [Anthropic's stock market debut this week](/article/anthropic-prices-ipo-83-share-1-5-trillion-valuation) was one of the biggest on record.
The incidents show the trade-off built into this new wave of AI agents. To do useful work on a person's behalf, an agent needs deep access to their phone or computer. That same access is what lets it read messages from apps the user never connected, or act on conversations that were never theirs. The same risks have shown up elsewhere: [AI agents that get hacked can leave customers liable](/article/banks-ai-agents-hacked-customers-liable), and some banks have responded by blocking agents outright.
Reuters could not establish whether Muse was showing users other people's real bank balances or inventing the numbers. Meta spokesperson Daniel Roberts called the examples "isolated incidents" and said many of the flagged messages were hallucinations, the AI term for made-up output, that had already been fixed. He said Muse can only reach other apps through permissions users explicitly grant.
But tests run by Reuters with staff at Signal and the AI Now Institute found the access was real and repeatable. On an Android phone, Muse used its permission to read push notifications to pull in Signal message text. On a Mac, it used the computer's accessibility and screenshot tools. The same requests were blocked on an iPhone.
What happens next
Apple has already moved. It changed macOS this week to give people more control over what AI agents can access, sharply restricting developers' use of accessibility tools and adding a new set of agent-specific permissions. The change followed a column by an Inc. Magazine writer who accused Muse of reading private chats from the Messages app on his Mac. "We are committed to ensuring users understand exactly what they are allowing, so they can make informed choices about their data and privacy," Apple said in a statement.
Google said it was considering similar changes, while noting that Android users "must deliberately grant permissions" for apps to read notifications. Signal president Meredith Whittaker accused Meta of launching Muse "without showing any respect for basic privacy or cybersecurity," calling the deployment of AI agents with wide device access "technologically reckless." She called for app makers to get consistent technical ways to mark their apps off-limits to agents, and for operating system providers to enforce those boundaries.
Meta warns new users that Muse may make mistakes, and its terms of service hold users responsible for the agent's actions. Meta did not immediately respond to TechCrunch's request for comment.



