Meta employees say Muse AI agent mixed up accounts and exposed bank balances, Reuters reports
Meta says employees test different versions of Muse than the public gets. Muse launched in September and has passed 10 million downloads.
What happened
Meta has an AI agent called Muse. Reuters says Meta employees posted internal complaints that it mixed up conversations between different accounts. An AI agent is software that acts for you inside apps and chats.
Why it matters to you
If an agent shows one person's private data to another, users could see bank balances or messages that are not theirs. It also raises the question of what apps the agent can reach.
The takeaway
Meta says the reports come from employee testing versions, not the public app.
What it means for you
The reports describe cross-account data leakage and apparent access to other apps, two failure modes to test for in any agent that touches personal accounts.
Companies considering agents that connect to messaging and finance apps have a concrete example of the privacy questions to ask vendors about permissions and data isolation.
Reuters reports internal privacy complaints about a Meta product that launched in September and has more than 10 million downloads; Meta says the issues involve employee test versions.