Every popular product attracts copycats, and personal AI agents are now popular. Meta's Muse passed 10 million US downloads in its first 40 days, Claude Duet launched behind a waitlist, and OpenAI's dots need ChatGPT Pro (from $100 a month) or Business Premium. Those three facts (huge demand, limited access, high prices) are exactly what impostor apps, fake browser extensions and "early access" invite scams exploit.

We have not found a published count of fake agent apps as of November 2026. The guidance below comes from how each vendor actually distributes its agent, documented supply-chain incidents such as ClawHavoc, and long-standing FTC advice on app and phishing scams.

Where do the real agents actually come from?

Knowing the genuine route is the fastest way to spot a fake. As of November 2026:

Agent Where to get it What a real sign-up involves
Meta Muse Muse app (iOS, Android) or muse.ai Meta account; 18+ ("or the age of majority"); free tier plus $20 and $100 plans (Meta)
OpenAI dots Inside ChatGPT; Slack and Teams for work ChatGPT Pro ($100, $200 or $500 a month) or Business Premium; no separate dots app
Claude Duet claude.com product page waitlist; standalone iOS and Android apps; claude.ai US Claude account, eligible plan and an invitation (Anthropic)
Grok Bot Linked from xAI's docs at docs.x.ai xAI/Cursor account; trials via the iOS App Store convert to paid unless cancelled (xAI)
OpenClaw github.com/openclaw/openclaw and docs.openclaw.ai Free, MIT-licensed; installed with npm; no official paid app

For Android, Anthropic's Claude Duet app is listed on Google Play under the package name com.anthropic.aibo, separate from the main Claude app. Checking the package name or developer name is a quick way to tell an official listing from a lookalike.

What forms do the fakes take?

Lookalike apps

An app with a similar name, a familiar icon and a vague developer name, often promising the agent in a country where it is not available yet. Most of these agents are US-only or limited to a few countries, so "Muse for the UK" or "Duet for Android in Europe" from an unknown developer should raise questions. Our sister site explains why most AI agents are US-only.

Browser extensions

Extensions that claim to "add dots to Chrome" or "unlock Muse on desktop". An extension can read every page you visit, including email and banking. None of the big four agents is distributed as a third-party browser extension.

Fake invites and waitlist offers

Messages offering to skip a waitlist, often for a fee or in exchange for "verifying" your account. Anthropic says Duet invitations come from the waitlist or from an existing Duet user, and access is tied to the email you used to sign up. No vendor we checked sells waitlist places.

Fake skills and plugins

For self-hosted agents, the impostor can be a skill or MCP server. ClawHavoc planted hundreds of malicious skills on OpenClaw's ClawHub marketplace in early 2026, disguised as useful tools. See MCP server security risks.

"AI agent" money schemes

Apps and services promising an AI agent that will make you money. The FTC's Operation AI Comply (September 2024) targeted deceptive AI claims; then-Chair Lina Khan said "Using AI tools to trick, mislead, or defraud people is illegal."

What are the red flags?

Pros

  • Reached from a link on the vendor's own website
  • Connects accounts through the service's own consent screen
  • Developer and package name match the vendor
  • Pricing matches the vendor's published plans

Cons

  • Asks you to type a password for another service into a chat or form
  • Charges to skip a waitlist or to "activate" an invite
  • Developer name does not match Meta, OpenAI, Anthropic, xAI or the OpenClaw project
  • Promises the agent in a country the vendor has not launched in
  • Asks for payment by gift card, crypto or wire transfer

The first red flag matters most. Vendors say their agents avoid seeing your passwords: Claude Duet "never sees your passwords, browser cookies, sign-in tokens, or other credentials," Muse's agent "only ever sees a 'surrogate' token," and Grok Bot signs in through a vault, password manager or Secure Form "without seeing the values." An app that asks for your Gmail password in a chat box is not working the way the real products do. More on this in fake agent sign-in pages and passwords, 2FA and agent takeover.

How do you check before installing?

  1. Start from the vendorType the vendor's address yourself or use a bookmark, and follow its app-store links.
  2. Check the developerConfirm the developer or publisher name in the store listing matches the company.
  3. Compare pricingIf the price differs from the vendor's published plans, stop.
  4. Read the permissionsBe wary of extensions that want access to all websites or apps that want SMS and accessibility access without reason.
  5. Ignore unsolicited invitesGet waitlist access only through the vendor's own sign-up page.

What do you do if you installed a fake?

  1. Uninstall the app or extension.
  2. Change the password for every account you entered or connected, starting with email, and sign out other sessions.
  3. Turn on two-factor authentication where it is not already on.
  4. Review connected apps and revoke anything unfamiliar; see how to revoke AI agent access.
  5. If you gave card or bank details, call your card issuer or bank straight away.
  6. Report the scam to the FTC at reportfraud.ftc.gov, to the FBI's IC3 if you lost money, and to the app store or extension store that listed it.

Bottom line

The defence against impostor agents is boring and reliable: go to the vendor first, then to the app. Real agents connect your accounts through official consent screens, publish their prices and do not sell invites. Anything that asks for your passwords, payment to skip a queue or an unusual payment method is not the product it claims to be. Our AI agent safety checklist collects these checks in one place.