Internal chat logs from Meta, recently detailed by Reuters, indicate that the company’s own employees raised serious alarms regarding the privacy protections of the Muse AI agent. Staff members warned that the AI appeared to be accessing and reviewing encrypted private messages that users expected to remain confidential. According to the reports, the agent occasionally generated outputs based on information from private conversations that users had never intended to share with the tool.
The leaked communications suggest that Muse has struggled with maintaining strict conversation boundaries, at times mixing up data between different users. Technical staff expressed concern that the system could potentially expose sensitive personal information, highlighting a discrepancy between Meta’s public privacy commitments and the internal behavior of the AI during testing. These findings emerge as Meta continues a broad push to integrate generative AI features across its global messaging platforms.

