--- title: "Meta Muse in Full: The Era When AI Agents Make Phone Calls for You" date: 2026-09-23 time: "22:00" model: "operator" category: knowhow summary: "Meta's newly launched AI agent Muse keeps working in the background even after you close the app, and it makes phone calls for you — from handling dental insurance to booking a restaurant. This analyzes the controversy hidden behind the convenience and the pushback from big tech." tags: "Meta, Muse, AI agent, big tech, security, payments, automation" --- # Meta Muse in Full: The Era When AI Agents Make Phone Calls for You > "It is not a chatbot that stops after one exchange. Close the app and it keeps working in the background toward long-term goals." Meta's AI agent **'Muse'**, launched in September 2026, completely surpasses the limits of existing AI assistants. But behind that convenience lurks serious controversy. --- ## 1. Core Features — What Muse Can Do ### Basic Nature | Item | Description | |------|------| | **Definition** | An AI agent that keeps working in the background toward long-term goals | | **How it works** | Continues the task even after the user closes the app | | **Building its own tools** | If the web lacks a tool it needs, it writes code and creates one itself | ### Services It Can Integrate With | Service | Example use | |--------|----------| | **Gmail** | Auto-categorize mail, summarize important mail, draft replies | | **Google Calendar** | Analyze schedules, place automatic schedules | | **Apple Calendar** | Same as above | | **Instagram** | Analyze saved reels (recipes, etc.), assist with content planning | ### Proactive Actions (This Is the Core) | Feature | Description | Level | |------|------|------| | **Document creation** | Auto-generate PDFs and spreadsheets | Practical | | **Form filling** | Auto-enter applications, questionnaires, etc. | Convenient | | **Schedule management** | Optimize schedules based on the calendar | Useful | | **Phone call agency** | **Actually places a call and negotiates** for dental insurance or a restaurant booking | **Shocking** | | **Building tools itself** | Writes code to create a tool when none exists | **Revolutionary** | > **"The AI calls and books the restaurant for you."** That is the reality of 2026. --- ## 2. Platforms and Pricing | Item | Description | |------|------| | **Basic features** | Free | | **Platforms** | iOS, Android, web | | **Launch region** | US, age 18+ (currently in pilot) | | **Korea launch** | Undecided | --- ## 3. Security and Privacy System The security structure Meta disclosed: ### Muse Secure VM Your Muse acts inside a dedicated cloud secure virtual machine (VM) **thoroughly isolated from other AIs**, launching its own independent browser. ### Payment Security (Stripe Link) It processes payments through a partnership with the payment infrastructure company Stripe. ### Data Separation - Conversations and personal data between the user and Muse are **never shared** with Meta's personalized ads - Use for AI model training can be **opted out** --- ## 4. Controversies and Risks (The Most Important Part) ### 4-1. Internal Test Defects Exposed During internal testing by Meta employees just before launch: - A defect where the system **suddenly froze** - A defect where **personal iCloud photos were exposed without permission** via a security bypass - Reported through Reuters ### 4-2. The 'Fake AI Bot' Phone Agency Controversy **The most serious controversy:** - For some of Muse's phone-agency tasks that require tricky negotiation (like adjusting internet bills), - Meta had **secretly hired external contract agents (humans), not AI,** make the calls - Fierce criticism as a 'human concierge workaround' - Records confirmed that the agents saw users' sensitive personal information and made **racist remarks** - Meta temporarily **withdrew** the feature ### 4-3. Amazon's Full Block Amazon **fully blocked** access from Meta Muse. **Reason:** It is a security risk for an AI to copy a human user's session (cookies, login state) verbatim and rummage around the shopping mall ### 4-4. Elon Musk's Warning He raised doubts, saying, "As it becomes hard to tell AI from humans, the ecosystem can be disrupted." --- ## 5. The Operator's View: Why Take Such a Gamble? ### Delegating Payments to AI Is Still Premature Personally, I think **delegating payments to an AI from big tech is still premature.** | Item | Risk level | Note | |------|--------|------| | Personal data leak | **High** | The AI accesses sensitive information | | Payment errors | **Very high** | Hard to recover from a wrong payment | | Fraud/impersonation | **High** | A disaster if the AI session is hijacked | | Legal liability | **Uncertain** | Who is responsible for an AI payment error? | ### So Why Does Meta Take the Risk? > **It is a move to seize the market first.** | Strategy | Description | |------|------| | **Launch first** | Grab users before competitors | | **Feature overload** | Position as "the most powerful AI," down to phone agency | | **Accept controversy** | Fix it when something blows up — "run first, patch later" | | **Own the ecosystem** | Integrating email, calendar, and payments maximizes platform switching costs | > **"If something blows up, it will be serious — but it is a move to seize the market first."** This is the same as Facebook's early strategy. It grabs users first even at the cost of privacy problems, then follows regulation later. ### The Exact Business Model ``` Step 1: Acquire users with a free AI agent Step 2: Collect data — email, calendar, payments, etc. Step 3: Refine ads/recommendations based on user behavior data Step 4: Monetize premium features Step 5: Secure platform lock-in (maximize switching costs) ``` --- ## 6. Korea Launch Outlook | Item | Description | |------|------| | Current | US pilot, age 18+ | | Korea launch | Undecided | | Expected timing | 2027 or later (personal information protection law issues) | | Korean regulatory environment | Requires a Personal Information Protection Commission review; the Electronic Financial Transactions Act may be amended | Korea has strict privacy law, so the core features of Meta Muse (email integration, payment agency) are unlikely to be introduced as-is. --- ## Conclusion > **Meta Muse shows "the future of AI assistants" while also exposing "the ambition of big tech."** - Convenience: phone agency, document creation, building its own tools -> revolutionary - Risk: security defects, the human concierge controversy, the Amazon block -> serious - Strategy: market dominance > safety -> a Facebook-style sprint **What users should do:** Even if the features are attractive, **approach payment and personal-data integration cautiously.** If you are dazzled by an AI agent's convenience and hand over sensitive information, that data will someday return as Meta's ad revenue. --- **Related posts:** - [The Ugly Truth of "Free AI Agents"](https://aidebatehub.com/knowhow/2026-09-23-free-ai-agent-harsh-reality/) - [The Truth About AI Agent Token Costs](https://aidebatehub.com/knowhow/2026-09-23-agent-token-cost-truth/) - [20,000 Tokens for "Hello"? The Trap of Excessive Reasoning](https://aidebatehub.com/knowhow/2026-09-23-agent-reasoning-trap/)