Meta’s AI agent Muse mixes up conversations, exposing private data, report says
agents meta
| Source: TechCrunch | Original article
Meta's AI agent Muse is mixing up conversations, surfacing unrelated childcare requests, career advice, product orders and incorrect bank account details, exposing private data.
Meta’s internal AI assistant, Muse, is under fire after employees reported that the chatbot is cross‑linking conversations and leaking personal data. According to a Reuters investigation, staff who tested the service observed that Muse surfaced messages about childcare requests, career advice, product orders and even bank‑account balances that did not belong to the users involved. The mishap appears tied to a feature that lets users share Muse chats via web links; the links were discoverable without authentication, allowing anyone who stumbled upon them to view the content.
The incident matters because Muse is Meta’s flagship general‑purpose AI agent, positioned as a cornerstone of the company’s broader push into conversational AI. Privacy breaches of this scale undermine user trust and could invite regulatory scrutiny, especially in Europe where data‑protection rules are stringent. For a platform that already faces scrutiny over how its AI tools handle personal information, the flaw highlights lingering safety gaps in the rollout of agentic systems.
Meta has already paused the conversation‑sharing function while engineers work on a fix, but the episode raises questions about the timeline for a full public launch. Observers will be watching for a detailed remediation plan, any statements from Meta’s privacy office, and whether regulators will demand formal audits. The broader AI community is also likely to scrutinise safety protocols in other emerging agents, from coding assistants to autonomous business tools, as the industry grapples with balancing rapid innovation against robust privacy safeguards.
Sources
Back to AIPULSEN