Technology

Meta’s Muse AI Misfires, Pulling In Unrelated App Chats During Employee Tests

By

Meta’s Muse AI Misfires, Pulling In Unrelated App Chats During Employee Tests

During employee testing, Meta’s Muse AI agents mistakenly accessed Signal and other apps that had not been connected to them, and sent seemingly random, off‑topic messages. Reuters reports that Meta’s Daniel Roberts described these occurrences as “isolated incidents.” Some of the most concerning incidents involved cryptic, out‑of‑context messages that appeared to contain other people’s information, often phrased as replies to queries the users had never submitted.

Related stories