Urgent.News

What's breaking now, across thousands of outlets.

Tech

Meta's AI Agent Muse Is Mixing Up Conversations — and Showing People's Private Data to the Wrong Users

Imagine asking your AI assistant about the weather and getting, unprompted, someone else's bank balance in reply. Not your balance. Someone else's. That is what Meta employees say is happening inside Muse, the company's AI agent — and if the reports are right, it is the kind of bug that should make every developer building agentic AI stop and think. A Reuters investigation published November 18,…

Meta's AI agent, Muse, has been mixing up conversations and showing private data to the wrong users. Reports from November 18, 2026, and confirmed by TechCrunch on November 19, 2026, detail instances where Muse displayed unrelated bank balances, childcare requests, career advice, and even product orders to users. For example, an employee recounted Muse providing their Capital One balance and Charles Schwab balance when discussing weather and tides.

Another user received a reminder on their personal phone to reply to a message on Signal, an end-to-end encrypted app, despite Muse having no access to such data. These incidents, described as "isolated" by Meta, highlight a significant failure in the agent's ability to maintain user identity and ensure data privacy. While Meta claims these were hallucinations and have been fixed, Reuters could not verify if the presented balances were real or fabricated.

The Signal case further raises concerns, as the agent appears to cross boundaries between accounts and apps, suggesting more structural issues. The story underscores the critical importance of accurate identity matching and robust access control in agentic AI systems, where even minor bugs could lead to privacy breaches with severe implications.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in Tech

Extending the UHPCM Patch Monitor Parser to Handle “Computer Name” Headers, Header‑less Pages, and Reboots Without an OC Column

Extending the UHPCM Patch Monitor Parser to Handle “Computer Name” Headers, Header‑less Pages, and Reboots Without an OC Column TL;DR: I refactored lib/parser.mjs to accept a single‑line “Computer…

  • UHPCM patch monitor parser extended to handle "Computer Name" headers.
  • Parser now accommodates pages without headers and reboots without OC column.
  • Changes ensure robust parsing across varied input formats.

The VMware vCenter Zero-Day: A Management-Plane Breach and the Cost of Waiting

When the management plane is owned, the estate is suspect Most vulnerability stories describe a bug that was quietly patched and then forgotten.

  • Chinese APT group exploited VMware vCenter zero-day vulnerability before patch release
  • Remote code execution (RCE) allowed without valid credentials, bypassing authentication
  • Immediate patching and isolation crucial to mitigate attack impact

More from Thursday 19 November →