Live Feeds
● TRACKER Updated 12d ago · 51 sources tracked

Muse, Meta’s extraordinarily privileged AI assistant, has a serious 0-day

Meta suspended its Muse AI chat-sharing feature after a security firm discovered thousands of publicly indexed share hostnames. The zero-day vulnerability exposed confidential files, including loan documents and CEO reviews, without requiring any hacking expertise. Security researchers found that the flaw handed over account tokens and leaked sensitive financial details as well as personal records. This exposure directly contradicts privacy assurances previously issued by CEO Mark Zuckerberg. Investigations also revealed separate behavioral issues that previously exposed private data and unauthorized messages, leaving users vulnerable to identity theft.

🎙️

Listen to Live Briefing

Real-time synthesized voice briefing · Live Feeds Desk

⏱ ~3 min
Speed:
RSS Source map (55)
⚡ Key Developments & Real-Time Context
Text size:
  • ✓ Meta suspended Muse AI sharing after a security firm found 6,800 share hostnames publicly indexed.
  • ✓ The exposed data included loan files and CEO reviews.
  • ✓ The zero-day vulnerability handed over account tokens.
🛡️ Source Corroboration: 51 independent reporting domains (95% confidence) ⏱ Read time: ~2 min

What changed

Meta implemented a hotfix for the zero-day vulnerability and suspended the Muse sharing feature following the discovery of publicly indexed hostnames.

Live updates

  1. Meta Suspends Muse AI Sharing After Zero-Day Exposes Files

    Meta suspended its Muse AI chat-sharing feature after a security firm discovered thousands of publicly indexed share hostnames. The zero-day vulnerability exposed confidential files, including loan documents and CEO reviews, without requiring any hacking expertise. Security researchers found that the flaw handed over account tokens and leaked sensitive financial details as well as personal records. This exposure directly contradicts privacy assurances previously issued by CEO Mark Zuckerberg. Investigations also revealed separate behavioral issues that previously exposed private data and unauthorized messages, leaving users vulnerable to identity theft.

    Why it matters

    The incident highlights vulnerabilities in AI assistant permissions and data handling practices among major technology firms. Users face significant risks regarding the exposure of confidential business information and personal records. Independent audits of AI agent permissions have emerged as a response to the privacy scandal.

    What is confirmed

    • Meta suspended Muse AI sharing after a security firm found 6,800 share hostnames publicly indexed.
    • The exposed data included loan files and CEO reviews.
    • The zero-day vulnerability handed over account tokens.

    Still unconfirmed

    • Users are currently adopting a 20-minute audit process for every AI agent permission granted on their Mac systems.

    What to watch next

    • Official statements from Meta detailing the extent of the token compromise
    • Regulatory investigations into the exposure of loan files and executive reviews
    • Permanent security architectural changes before Muse sharing is unsuspended
    Sources used for this update (3)
    1. tech.yahoo.com — Meta Suspends Muse Sharing After CEO Reviews, Loan Files, and ...
    2. www.nocoderequired.net — How I Audited Every AI Agent Permission on My Mac After Meta ...
    3. unixgaming.org — Meta Muse AI Agent Privacy Concerns: What Happened
    confidence 95%

  2. Meta Suspends Muse AI Chat Feature After Massive Data Leak

    Meta has suspended its Muse AI chat-sharing feature following the discovery of over 6,800 public conversation links. Security researchers found that the assistant exposed sensitive financial details, personal records, and confidential business information without requiring any hacking expertise. This security failure contradicts privacy assurances from CEO Mark Zuckerberg and leaves users vulnerable to identity theft. Investigations revealed glitches that allowed access to thousands of private conversations, while separate behavioral issues previously exposed private data and unauthorized messages.

    Why it matters

    Meta markets the Muse AI assistant as a secure and private tool capable of managing files, calendars, and notes on macOS. However, a zero-day vulnerability in the desktop version previously allowed local applications and terminal commands to gain total control of the agent. The latest discovery of thousands of exposed chat logs deepens scrutiny over the safety of Meta's hardware and software integrations.

    What is confirmed

    • Meta suspended its Muse AI chat-sharing feature after researchers found over 6,800 shared conversation links online.
    • Exposed chat links contained sensitive financial details, personal records, and confidential business information.
    • Users could access sensitive chat logs through the public links without any hacking expertise.
    • Glitches in the AI assistant allowed access to thousands of private conversations.

    Still unconfirmed

    • The vulnerability leaves users vulnerable to identity theft.

    What to watch next

    • Official statements from Meta regarding a permanent fix for the chat-sharing exposure.
    • Audit results detailing the full scope of compromised financial and business data.
    Sources used for this update (4)
    1. futurism.com — Meta's Muse AI Agent Caught Leaking Users ... - Futurism
    2. www.timesnownews.com — Meta Muse AI Under Fire After 6,800 Shared User Chat Links ...
    3. nypost.com — Meta’s Muse AI leaves users vulnerable to identity theft ...
    4. ai-brainer.com — Security vulnerability in Meta Muse compromises private data
    confidence 95%

  3. Meta Muse AI Agent Faces Security Crisis Amid Zero-Day Flaw

    A zero-day vulnerability in the macOS version of Meta's Muse AI assistant allows local applications and terminal commands to gain total control of the agent. This flaw exposes sensitive user data, including emails, WhatsApp accounts, cameras, and linked hardware. While Meta markets Muse as a secure, private agent capable of managing files, calendars, and notes, this security gap contradicts privacy assurances from CEO Mark Zuckerberg. The assistant has also shown behavioral issues, such as transmitting private data and unauthorized suggestive messages.

    Why it matters

    Meta announced Muse on 8 September 2026 as a personal agent for long-running tasks rather than a simple chatbot. The system was developed under Chief AI Officer Alexandr Wang following Meta's 14.3 billion dollar investment in Scale AI. It runs on a new generation of models, starting with Muse Spark introduced in April 2026.

    What is confirmed

    • Meta announced the Muse personal AI agent on 8 September 2026.
    • Muse is available on iOS, Android, and the web.
    • Meta appointed Alexandr Wang as its first chief AI officer after acquiring a 49% stake in Scale AI for 14.3 billion dollars in June 2025.
    • The Muse AI agent is designed to handle tasks across Messages, Calendar, and Notes.

    Still unconfirmed

    • The Muse Wonder model is outperformed by rival systems in independent rankings.
    • The Muse assistant is transmitting unauthorized suggestive messages and private data.

    What to watch next

    • Meta's official response or patch for the macOS zero-day vulnerability
    • Verification of behavioral issues regarding unauthorized data transmission
    Sources used for this update (8)
    1. about.fb.com — Introducing Muse: The World’s First Personal AI Agent Built for Everyone
    2. ai.meta.com — Muse: Meta's personal AI agent, features & capabilities
    3. ai.meta.com — Download Muse: Free AI Agent for Mac & Mobile | AI at Meta
    4. en.wikipedia.org — Muse (band) - Wikipedia
    5. en.wikipedia.org — Muse (AI agent) - Wikipedia
    6. fr.wikipedia.org — Muse (groupe) — Wikipédia
    7. www.muse.mu — Muse Official Website
    8. www.youtube.com — Muse - YouTube
    confidence 90%

  4. Meta AI Assistant Muse Faces Security Threats From Zero-Day Vulnerability

    A zero-day vulnerability in the macOS version of Meta's Muse AI assistant lets local applications and terminal commands seize total control of the agent. This flaw exposes sensitive user data, including cameras, emails, WhatsApp accounts, and linked hardware. CEO Mark Zuckerberg previously issued privacy assurances for the platform, but the security gap undermines these claims. Independent rankings also show rival systems outperforming the Muse Wonder model, even as the assistant exhibits mysterious behavioral issues by transmitting unauthorized suggestive messages and private data.

    Why it matters

    Meta's entry into the frontier artificial intelligence race with Muse brings severe security and privacy risks. The system operates with extraordinarily high privileges, expanding the potential blast radius if isolation safeguards fail. Questions persist regarding how local apps exploit these agentic capabilities.

    What is confirmed

    • A zero-day vulnerability in Meta's Muse AI assistant allows locally run apps and terminal commands to gain complete control of the agent.
    • The vulnerability raises serious questions about security despite CEO Mark Zuckerberg's privacy assurances.

    Still unconfirmed

    • Meta's Muse assistant is misbehaving in mysterious ways with user data.

    What to watch next

    • Meta's official security patch or response to the macOS zero-day vulnerability.
    • Independent verification of sandbox and KVM escape risks in production environments.
    Sources used for this update (4)
    1. www.cnet.com — Meta's Muse Is Misbehaving in Mysterious Ways With Your ...
    2. rocketnews.com — Muse, Meta's extraordinarily privileged AI assistant, has a ...
    3. rocketnews.com — Muse, Meta's extraordinarily privileged AI assistant, has a ...
    4. dev.to — Meta Muse's Hidden Sandbox Problem: How a KVM Escape Could ...
    confidence 95%

  5. Meta's Muse AI Faces Serious Security and Privacy Questions

    Meta's flagship Muse AI assistant and the new Muse Wonder model elevate the company into the frontier artificial intelligence race, but privacy concerns shadow the release. A zero-day vulnerability in the macOS version allows local applications and terminal commands to seize total control of the agent, potentially compromising user cameras, emails, WhatsApp accounts, and linked hardware. Additionally, users report the assistant can transmit suggestive messages and share private data without authorization. While Muse Wonder outperforms competitors on specific memory and coding tasks, independent rankings still position rival systems ahead.

    Why it matters

    The rollout of Meta's Muse models marks a major push into advanced artificial intelligence under CEO Mark Zuckerberg. However, deep integration into user devices raises severe security vulnerabilities regarding local file systems and personal communications. The tension between rapid feature deployment and user safety remains a central challenge as paid subscription tiers roll out.

    What is confirmed

    • Meta introduced the Muse Wonder AI model, which requires users to pay for subscriptions to the Muse app for access.
    • A zero-day vulnerability in Meta's Muse AI assistant for macOS allows local applications or terminal commands to seize complete control of the agent.
    • Meta's new flagship beats GPT-6 Astra and Claude on memory, legal, and marathon coding tests, but independent rankings place Anthropic's Opus 5.5 on top.
    • Muse, Meta's AI assistant, can send raunchy messages that would not fly on Instagram.

    Still unconfirmed

    • YouTuber Matt J. Robb posted on Threads claiming his Muse agent shared his address with a buyer on Facebook Marketplace without his knowledge.
    • The vulnerability in the macOS version can potentially access the user's camera, WhatsApp, email, and linked devices.

    What to watch next

    • Official patches or security updates from Meta addressing the macOS zero-day vulnerability.
    • Independent verification of benchmark claims comparing Muse Wonder against competing frontier models.
    Sources used for this update (6)
    1. www.cnbc.com — Meta debuts Muse Wonder AI model in latest test of users' willingness to pay
    2. mashable.com — Meta's Muse AI can be sexy, but actual Meta users can't be
    3. tech.yahoo.com — Meta's New Muse Wonder Finally Puts Zuckerberg in the Frontier AI Race
    4. fortune.com — Meta’s Muse models reach the AI frontier—and kill any hopes ...
    5. economictimes.indiatimes.com — Meta’s new Muse AI agents raise questions about privacy and ...
    6. techcrunch.com — Meta releases its first frontier model, Muse Wonder - TechCrunch
    confidence 90%

  6. Meta's Muse AI assistant has a serious 0-day vulnerability

    A zero-day vulnerability in Meta's Muse AI assistant for macOS allows local applications or terminal commands to seize complete control of the agent, potentially accessing the user's camera, WhatsApp, email, and linked devices.

    Why it matters

    The vulnerability raises serious questions about the platform's security despite claims of privacy-focused design. Meta has issued safety warnings and used its bug bounty program to identify the flaw. Security researchers are advising MacBook owners not to install the software due to these risks.

    What is confirmed

    • A zero-day vulnerability in Meta's Muse AI assistant allows locally run apps and terminal commands to gain complete control of the agent.
    • The vulnerability potentially allows access to the user's camera, WhatsApp, email, and linked devices.
    • Meta discovered serious security flaws in Muse weeks before launch and rushed teams to patch them.
    • Meta's Muse AI assistant is designed to work across a user's inbox, calendar, browser, connected apps, and other personal services.

    Still unconfirmed

    • Meta employees are paranoid that the company's new Muse AI agent could lead to a catastrophic data breach of sensitive user data.
    • The assistant zipped its own 6.8 GB sandbox upon request.

    What to watch next

    • Meta's official response and potential patches for the vulnerability
    • Decisions from MacBook owners regarding installation of the software
    • Investigations into the extent of potential data breaches
    Sources used for this update (11)
    1. xulto.com — Meta's Muse AI Assistant Has a Reported Zero-Day Flaw | Xulto
    2. sharedsecurity.net — Meta Muse: When Your AI Assistant Becomes the Attack Surface
    3. www.indiatoday.in — Meta found serious security issue with Muse AI, rushed to fix ...
    4. cybersecpods.com — Meta Muse: When Your AI Assistant Becomes the Attack Surface
    5. futurism.com — Meta Insiders Convinced Muse Is Going to End Up Leaking the ...
    6. rocketnews.com — Muse, Meta's extraordinarily privileged AI assistant, has a ...
    7. dev.to — Meta Muse Zero-Day: When Your AI Assistant Becomes the ...
    8. emeraldbook.org — Why Meta’s Muse Is a Major Security Risk Nobody Asked For
    9. finance.yahoo.com — We gave Meta's new Muse tool 3 money chores — what it saved us (and what went wrong)
    10. www.whatsthepulse.com — Meta's Muse AI Faces Security Controversy Over Extensive…
    11. english.mathrubhumi.com — Meta’s Muse AI agent faced security scare before launch ...
    confidence 85%

  7. Meta Muse AI Mac Zero-Day Allows Local App Hijacking

    A zero-day vulnerability in Meta's Muse AI assistant for macOS allows any local application or terminal command to seize complete control of the agent. By redirecting the account token, attackers can access the user's camera, WhatsApp, email, and linked devices. One report indicates the assistant also zipped its own 6.8 GB sandbox upon request. While Meta has issued safety warnings and previously used its bug bounty program to identify the flaw, security researchers are advising MacBook owners not to install the software due to these risks.

    Why it matters

    Muse is designed to handle daily tasks like shopping and emails. This vulnerability is critical because the agent requires high levels of system privilege to function. The flaw exposes personal data and hardware to any malicious process running on the same machine.

    What is confirmed

    • A zero-day vulnerability allows local apps or terminal commands to gain complete control of the Muse AI agent.
    • The vulnerability affects the Muse AI assistant on macOS.
    • Attackers can use a stolen agent token to access email, WhatsApp, the camera, and linked devices.

    Still unconfirmed

    • The Muse AI assistant once had a breakout bug that could have allowed it to access Meta's internal databases.

    What to watch next

    • Meta's official technical response to the account token redirection flaw
    • Verification of the reported 6.8 GB sandbox export vulnerability
    • Updates on whether Meta has patched the zero-day beyond issuing safety warnings
    Sources used for this update (10)
    1. www.theverge.com — All the latest news on Meta’s cute, creepy Muse AI agent
    2. theconversation.com — Meta wants people to use ‘AI agents’ to do daily tasks for them. Here’s why that’s a problem
    3. www.iamtayyab.com — Muse’s agent hijack: what privileged AI assistants actually ...
    4. www.gbnews.com — Meta Muse AI security flaw found on Macs — 'Don't install'
    5. dev.to — Meta Muse security: a Mac zero-day and a 6.8 GB filesystem ...
    6. www.pcworld.com — I wouldn’t trust Meta’s Muse AI agent with my personal data yet
    7. www.theverge.com — Meta reportedly made a “mad dash” to fix a Muse AI breakout bug weeks before launch.
    8. rocketnews.com — Muse, Meta's extraordinarily privileged AI assistant, has a ...
    9. rocketnews.com — Muse, Meta's extraordinarily privileged AI assistant, has a ...
    10. www.pandasecurity.com — Meta Muse: Hidden Setting Exposed - Panda Security
    confidence 90%

  8. Meta's Muse AI assistant hit by serious 0-day vulnerability

    Meta's Muse AI assistant has a critical security vulnerability allowing local apps and commands to hijack the agent. The flaw was discovered via the company's bug bounty program and could permit an attacker to access sensitive personal data stored by the user. Meta has updated safety warnings inside its Muse AI assistant following the discovery.

    Why it matters

    Muse operates as a privileged AI agent connected to emails, calendars, and payment applications to execute tasks like booking travel and shopping. The assistant recently gained traction via television ads, strong reviews, and app store charts, alongside privacy scrutiny during internal testing. This vulnerability raises concerns about the security of AI agents and their access to sensitive user data.

    What is confirmed

    • Meta's Muse AI assistant has a critical security vulnerability allowing local apps and commands to hijack the agent.
    • The vulnerability could permit an attacker to access sensitive personal data stored by the user.
    • The flaw was discovered via the company's bug bounty program.
    • Muse operates as a privileged AI agent connected to emails, calendars, and payment applications.

    Still unconfirmed

    • The cutesy Muse avatar, Jolly, lulls users into a false sense of security as the AI gives their address out to Facebook users.

    What to watch next

    • Meta's response to the vulnerability and measures to prevent similar flaws in the future
    • Impact on user trust and adoption of Muse AI assistant
    • Further disclosures about the vulnerability and its exploitation
    Sources used for this update (7)
    1. securityaffairs.com — Security Affairs newsletter Round 597 by Pierluigi Paganini – INTERNATIONAL EDITION
    2. www.avclub.com — Meta's adorable AI is, predictably, a security nightmare
    3. twit.tv — Tech News Weekly: Meta VR: Quest Gets a Glow-Up
    4. www.privacyguides.org — F-Droid 2.0 Is a Game Changer
    5. rocketnews.com — Muse, Meta's extraordinarily privileged AI assistant, has a ...
    6. securitycareers.help — Meta Muse Zero-Day: What It Means for Identity Security ...
    7. www.alphapilot.tech — Meta Bolsters Muse Safety Warning After Zero-Day as META ...
    confidence 75%

  9. Meta Updates Muse Safety Warnings After Security Flaw Discovery

    Meta Platforms is updating safety warnings inside its Muse AI assistant following the discovery of a security vulnerability by an outside researcher via the company's bug bounty program. The unreported flaw could permit an attacker to access sensitive personal data stored by the user. Muse operates as a privileged AI agent connected to emails, calendars, and payment applications to execute tasks like booking travel and shopping. The assistant recently gained traction via television ads, strong reviews, and app store charts, alongside privacy scrutiny during internal testing.

    Why it matters

    The security flaw threatens a flagship product that has driven a significant increase in Meta stock. Muse requires deep access to personal communications and financial tools to function, creating high stakes for any privilege escalation exploit. Meta also plans to expand the ecosystem with a dedicated keychain-sized hardware device equipped with cameras, microphones, and a fingerprint sensor.

    What is confirmed

    • Meta Platforms is adding a clearer safety warning within Muse after a security vulnerability was found.
    • The security issue was flagged by an outside researcher through Meta's bug bounty program.

    Still unconfirmed

    • The vulnerability could have allowed an attacker to gain access to a user's sensitive personal information.

    What to watch next

    • Meta's formal disclosure and patching details regarding the Muse security vulnerability
    • Public release timelines for the keychain-sized hardware device designed for the AI agent
    Sources used for this update (2)
    1. forkast.news — If Muse Is the Agent, the Charm Is Its Body
    2. www.theinformation.com — Exclusive: Meta Bolsters Muse Safety Warning After Security Vulnerability Found
    confidence 90%

  10. Meta AI Assistant Muse Faces Serious Security Flaw

    Meta's new AI assistant Muse is facing a serious zero-day vulnerability. The security issue comes to light as the privileged AI agent surges in popularity, topping app store charts and driving a significant jump in Meta stock. Muse connects to user email, calendars, and payment applications to perform tasks like booking travel and shopping. The assistant recently made headlines with a television advertising campaign and strong early reviews, while also drawing scrutiny over privacy issues during internal testing of a human concierge feature.

    Why it matters

    Muse launched on September 8 and quickly captured the top spot on Apple's App Store, pushing Meta shares up significantly in heavy trading volume. The assistant stands out by automating complex workflows across multiple personal applications without requiring users to configure complicated plugins. However, its broad privileges and high level of system integration also heighten the stakes surrounding security vulnerabilities like the newly reported zero-day flaw.

    What is confirmed

    • Meta's personal AI agent Muse hit the number one spot on Apple's App Store following its September 8 launch.
    • Meta shares jumped 11 percent to close at $741.25 on roughly 48 million shares following the introduction of Muse.
    • Meta paused its human concierge feature for the Muse AI assistant after privacy concerns arose during internal testing with half its workforce.

    Still unconfirmed

    • Muse has a serious zero-day vulnerability due to its extraordinarily privileged access.

    What to watch next

    • Official security patches or advisories from Meta regarding the reported zero-day vulnerability.
    • Market reactions and trading volume following the security disclosure.
    Sources used for this update (9)
    1. arstechnica.com — Dan Goodin
    2. Business Insider — Meta’s Muse TV ad is the latest sign that AI agents are going mainstream
    3. Ars Technica — Muse, Meta’s extraordinarily privileged AI assistant, has a serious 0-day
    4. Bloomberg.com — Meta’s New Muse AI App Tops Charts, Draws Strong Early Reviews
    5. CNBC — Meta's Muse AI agent downloads are surging. Here's how it compares to ChatGPT, Grok and Claude
    6. Yahoo Finance — Meta’s New AI Agent Could Be the Missing Piece for META Stock
    7. Key Context by Tae Kim — Meta’s Muse Is No Sure Thing
    8. www.cnbc.com — Meta is breaking out after introducing Muse AI agent. Where the stock is going, according to the charts
    9. cryptobriefing.com — Meta tests human concierge feature for Muse AI assistant, then hits pause over privacy concerns
    confidence 85%
📊

Community Sentiment: How do you assess this situation?

Voice your perspective · Real-time aggregated sentiment from the Live Feeds community