AI News/AI News

Meta employees flag Muse agent mixing up users' conversations and private data

A Reuters report says internal Meta posts describe Muse surfacing other accounts' messages and bank balances, a warning for teams adopting AI agents.

Key points

  • Reuters, reviewing internal Meta posts, reports employees saw Muse mix up conversations across accounts and surface other people's private data.
  • Examples included unrequested messages about childcare and orders, plus bank balances an employee said were not theirs.
  • Muse launched in September and has passed 10 million downloads, per TechCrunch's summary of the report.
  • Meta did not immediately respond to TechCrunch's request for comment.

Why it matters for marketers: Marketers piloting or building on consumer AI agents should treat data isolation and app permissions as a vendor-risk question, since a leak can quickly become a brand and compliance problem.

Meta employees have raised concerns that Muse, the company's AI agent, has been mixing up conversations across accounts and exposing private user data. The claims come from a Reuters report based on internal company posts, as summarized by TechCrunch.

According to that account, employees said Muse surfaced childcare requests, career advice, and product orders that they never made. One worker described an unprompted message in a weather and tide chat that included two bank balances. Another employee, using Muse on a personal phone, got a reminder to reply to her sister on Signal and wrote that she had not granted access to other apps.

Muse launched in September and has passed 10 million downloads, which makes it one of the fastest-growing apps on record, TechCrunch notes. Meta did not immediately respond to a request for comment, so the company's explanation and any fix are not yet known. The reports are employee accounts, not a confirmed finding.

For marketers and content teams, the story is a reminder that agents act across accounts and apps, and the permissions behind them matter as much as the output. If your team uses consumer agents for scheduling, messaging, or customer tasks, check what data they can reach and where conversations are separated.

Watch for a Meta statement, details on how many users were affected, and any changes to Muse's app permissions. Teams evaluating agent tools should ask vendors how they isolate user data before putting customer information anywhere near them.

Sources

This brief is our own summary. Read the full reporting at the source.

You read about AI automation. We run it.

DocFluence, built by Z Media, writes, designs, films and publishes for your brand every day, in your voice, from one login.