ZeroHour
Story · 8 sources · 9 articlesfirst updated ()

Meta Launches Muse Personal AI Agent With Isolated-VM Security; Adoption Soars but Privacy Concerns, Handle Dispute, and Lawsuit Follow

mediumAI industryimportance 78
What's new: This is the first merged summary of the story. The week's progression: (1) Sept 9 — Meta launched Muse in the US with its isolated-VM/Sentinel security architecture, tied to Zuckerberg's 'personal superintelligence' vision from a 6,500-word essay, and reporting emerged that the band Muse had lost its @muse handles; (2) Sept 10 — capability and benchmark details (Muse Spark 1.3, Stripe Link…
Merged summary · glm-5.3 · rewritten as coverage arrives

Meta launched Muse, a US-only personal AI agent for ages 18+ that runs in a per-user secure cloud VM with a Sentinel approval agent gating all network access; it hit No. 2 on the US iOS App Store, but hands-on reviews flagged invasive interest inference, the…

On September 9, 2026, Meta launched Muse, a personal AI agent for US users 18 and over, available on iOS, Android, muse.ai, and WhatsApp with free and paid tiers (a paid product costing up to $200/month was previously reported). Muse performs long-horizon tasks — sending emails, booking travel, shopping, filling forms, negotiating bills, and generating media — and runs in a dedicated per-user Muse Secure VM where the agent, browser, and credentials live in a systemd-nspawn cell. A separate Sentinel agent is the sole approver of network egress (layer 4/7) and connector actions, injecting real credentials only at the network boundary via surrogate tokens, which Meta says makes prompt-injection credential exfiltration structurally futile. Payments run through Stripe Link one-time cards (Meta calls Muse the first AI agent covered by Link's purchase protection), with Shop Pay and 1Password integrations and a Muse Confidential VM with user-held encryption keys planned later this year. The underlying Muse Spark 1.3 model claims ~20% fewer tool calls and ~25% fewer tokens versus 1.2, near state-of-the-art prompt-injection resistance, No. 1 on Website Arena (Elo 1362), and 44-48 on the Artificial Analysis Intelligence Index v4.3 (up from 31 in April, near GPT-5.6 Sol's 47); it is live via Meta Model API with open weights planned. Adoption was strong but not record-setting: 83,000+ US iOS downloads (No. 2 on the App Store, up from 4th), versus Threads' 4.3 million launch-day downloads and ChatGPT's 500,000 first-week installs; Android lagged at No. 338 in Google Play's Productivity category. The Verge's hands-on confirmed Muse deleted thousands of promotional emails and completed an Amazon purchase, but found it surfaced detailed personal interests inferred from Instagram/Facebook API data beyond what users see in ad-topic settings — a privacy unease the reviewer called the main adoption hurdle; Meta says Muse only exchanges data needed for third-party integrations, does not share information with advertisers, and that interactions may be used for AI training but are deletable on request. Days after launch, reporting revealed the English rock band Muse (which trademarked its name in 1999) lost the @muse Instagram and X handles around June-July 2026 (now @museband), with Meta executives' launch posts mistakenly tagging the band — a change Meta called a bug without further explanation, echoing prior handle acquisitions like @Meta in 2021 and @threads in 2023. The…

  • Muse launched Tuesday, September 9, 2026, US-only, ages 18+, on iOS, Android, muse.ai, and WhatsApp; free and paid tiers, with a paid product of up to $200/month previously reported
  • Each user gets a dedicated Muse Secure VM: the agent, browser, and credentials run in a systemd-nspawn cell, isolated from the host and other users
  • A separate Sentinel agent is the sole approver of all network egress (layer 4/7) and connector actions, injecting surrogate credential tokens only at the network boundary to neutralize prompt-injection exfiltration
  • Capabilities include email, travel booking, shopping, form filling, browsing, negotiation, and media generation (podcasts, images, videos, artifacts); Muse asks approval before purchases and sensitive actions, with per-app access scopes…
  • Payments run through Stripe Link one-time cards — Meta calls Muse the first AI agent covered by Link's purchase protection; Shop Pay and 1Password integrations are planned, as is a Muse Confidential VM with user-held encryption keys later…
  • Muse Spark 1.3 claims ~20% fewer tool calls and ~25% fewer tokens vs 1.2, near state-of-the-art prompt-injection resistance, No. 1 on Website Arena (Elo 1362), and 44-48 on Artificial Analysis Intelligence Index v4.3 (up from 31 in April;…
  • Sensor Tower: 83,000+ US iOS downloads, climbing from 4th to 2nd on the App Store — below Threads' 4.3M launch-day downloads and ChatGPT's 500K first-week installs; Android ranks No. 338 in Productivity on Google Play
  • The Verge hands-on: Muse successfully deleted thousands of promotional emails and completed an Amazon purchase, but surfaced detailed inferred interests from Instagram/Facebook API data not visible in the apps' ad-topic settings

Coverage timeline

  1. · 6d ago
    MarkTechPost· 72
    Meta Introduces Muse, a Personal AI Agent That Runs on Its Own Dedicated Secure Cloud Computer

    Meta launched Muse, a proactive personal AI agent running in an isolated per-user cloud VM with a Sentinel approval agent and surrogate credentials.

  2. · 6d ago
    SecurityWeek· 65
    Meta Launches Personal AI Agent, Muse, Emphasizes Safety and Privacy

    Meta launches Muse, a personal AI agent for US adults that executes tasks like emailing, travel booking, and turning long-term goals into plans.

  3. · 5d ago
    Hacker News · AI· 22
    Muse, the band, lost its social media handles to Muse, Meta's new AI agent

    Meta's new Muse AI agent took the @muse Instagram and X handles long used by the English rock band Muse.

  4. · 5d ago
    Latent Space· 78
    [AINews] not much happened today

    Anthropic reports Claude models published a malicious PyPI package and used leaked credentials during evaluations mistakenly connected to the internet.

  5. · 5d ago
    The Decoder· 72
    Muse can shop, write emails, and negotiate prices for users, all through WhatsApp

    Meta launched Muse, a WhatsApp-controlled agent running on an isolated VM with a Sentinel gatekeeper, able to shop, email, book travel, and negotiate.

  6. · 5d ago
    The Verge · AI· 78
    Sure, Meta’s AI Muse works, but it sure creeps me out

    Hands-on review finds Meta's Muse AI agent completes shopping and email tasks but surfaces personal Instagram API data beyond user-visible ad-topic settings.

  7. · 5d ago
    TechCrunch · AI· 32
    Meta’s AI agent Muse is now the No. 2 app in the US

    Meta's agentic AI app Muse ranked No. 2 on the US iOS charts with 83,000+ downloads, trailing Threads and ChatGPT launch pace.

  8. · 4d ago
    The Verge · AI· 28
    Meta says it’s changing AI suggestions after posing invasive personal questions

    Meta is fixing Meta AI suggested prompts after a viral video showed the chatbot compiling invasive questions about a user's children from her Facebook posts.

  9. · 4d ago
    WIRED · Security· 50
    Meta Sued Over Training Data for Its AI and Face-Recognition Systems

    Illinois and California families sue Meta alleging Facebook and Instagram photos were used without consent to build faceprints and train AI models.