---
title: You Can Say Yes to Muse – Just Not on Behalf of Your Friends
description: AI agents like Meta's Muse read your inbox, and everyone in it. Nine steps to protect yourself, your friends and your clients' data.
image: https://signal78.ai/hubfs/2026-10-21_asset_blog-friendsdata-robot_r01a.jpg
---

[Skip to content](https://signal78.ai/insights/you-can-say-yes-to-muse.-your-friends-cant#main)

[![Signal78 home](https://signal78.ai/hs-fs/hubfs/signal78-site/logo-signal78.png?width=95&height=30&name=logo-signal78.png)](https://signal78.ai/)

[Home](https://signal78.ai/) [Platform](https://signal78.ai/platform) [Services](https://signal78.ai/services) [About](https://signal78.ai/about) [Insights](https://signal78.ai/insights) [Book a demo](https://signal78.ai/insights/you-can-say-yes-to-muse.-your-friends-cant#demo)

[← Insights](https://signal78.ai/insights)

AI AgentsOctober 21, 20267 min read

# You Can Say Yes to Muse – Just Not on Behalf of Your Friends

![Matt Cyr](https://signal78.ai/hubfs/signal78-site/headshot-matt-cyr.jpg)**Matt Cyr** · Founder & President, Loop

![](https://signal78.ai/hubfs/2026-10-21_asset_blog-friendsdata-robot_r01a.jpg)

In this article

Share

<https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fsignal78.ai%2Finsights%2Fyou-can-say-yes-to-muse.-your-friends-cant>

## Meta’s agent is the loudest example, but it isn’t the only one asking for your inbox. Here are nine steps to protect yourself, the people in your network and the clients who trust you with their work.

In 2014, about 270,000 people took a personality quiz on Facebook. They agreed to share their data. Their friends didn’t, but the app pulled their friends’ data anyway. By the time anyone noticed, information on as many as 87 million people had ended up with Cambridge Analytica. Facebook later paid a record $5 billion penalty to the Federal Trade Commission over the privacy failures behind it.

Twelve years later, Meta has built something that makes that quiz look quaint. Muse, the AI agent with the fuzzy mascot [I wrote about last week](https://signal78.ai/insights/cute-agents-are-sweeping-something-under-the-rug), asks for your email, your messages, your contacts and your bank accounts. And as [privacy experts pointed out](https://www.theverge.com/tech/1011938/muse-ai-data-friends-consent), much of what lives in those places doesn’t belong to you. It belongs to your friends, your family, your clients and the plumber who sent you an invoice last spring.

Researchers who pulled Muse’s internal instructions found an hourly process that [builds and updates “a page for every person in the user’s life”](https://gizmodo.com/metas-muse-is-collecting-information-on-everyone-in-your-life-2000821697) — their birthdays, the topics that keep coming up, how close you are, even “what the relationship seems to need.” That includes people who have never touched Muse.

## We’ve Seen This Movie Before

Muse isn’t the only one. OpenAI’s Dots [gives each agent its own cloud computer](https://www.techspot.com/news/114043-openai-takes-meta-muse-always-dots-agents-less.html) and connects to more than 4,000 apps, and Instinct, a startup now [valued at $10 billion](https://www.ksl.com/article/51629454/ai-agent-firm-instinct-raises-1-billion-in-latest-funding-round), will call businesses and buy things on your behalf. Last fall, when AI browsers like Perplexity’s Comet and OpenAI’s Atlas arrived, I wrote that [AI was shifting the labor to machines and the privacy burden to humans](https://signal78.ai/insights/ai-is-shifting-the-labor-to-machines-and-the-privacy-burden-to-humans): by using them, we were consenting on behalf of people we might not even know.

A browser reads the page in front of you. An agent reads your whole life — your inbox, your texts, your calendar — and then acts on it. The burden just got a lot heavier.

So, what can you actually do? I’ll use Muse as the example, but every one of these applies to any agent you let into your accounts.

## Start With Your Own Settings

If you use Muse or any agent like it, these four steps take about 10 minutes.

**1. Turn off training.** Muse trains on your conversations by default. To stop it, open Settings, go to Data Controls and [turn off “Help improve our AI models.”](https://tech.yahoo.com/ai/meta-ai/articles/stop-meta-training-muse-data-163434020.html) Most agents and AI browsers have a similar switch, usually buried under data or privacy controls. Find it before you connect anything.

**2. Make it ask before it acts.** Muse is supposed to [check with you before sensitive actions](https://fastcompany.com/91603856/meta-wants-its-new-ai-agent-to-run-your-digital-life) like sending an email or making a purchase, and you can limit email to read-only. Do both, with any agent. Every time one offers to stop asking, say no. One user asked Muse to handle a Marketplace sale, and it [gave his home address to a stranger — then did it five more times](https://epic.org/metas-mass-data-collection-is-not-a-muse-ing/) after he told it to stop. Approvals matter even more now that [phishing emails can hide instructions](https://futurism.com/artificial-intelligence/ai-agent-sucker-phishing-emails) only your AI assistant can read. Keep a human on anything that sends, spends or shares.

**3. Ask what it knows about someone you love.** Muse keeps an audit trail and a memory, and it will tell you what’s in them if you ask. Pick a person — your sister, your business partner, your kid’s coach — and ask Muse to show you what it has on them. Then use the forget command on anything that isn’t yours to keep. It’s the fastest way I know to understand what you’ve actually agreed to.

**4. Check what your computer and browser are exposing.** On a Mac, agents that can read your messages usually need Full Disk Access. Apple is [tightening that permission because of AI agents](https://www.notebookcheck.net/Apple-is-tightening-Mac-Full-Disk-Access-over-AI-agents.1415473.0.html), but the change isn’t here yet. Until it is, open System Settings, then Privacy & Security, then Full Disk Access, and switch off anything that doesn’t need your entire computer. If you use an AI browser like Comet or Atlas, close the tabs holding your bank, your medical portal or a client’s files before you hand the assistant a task.

## Then Think About Everyone Else

Every account you connect is full of other people’s information, and they have no say in it.

**5. Connect less than you’re asked to.** Every agent will keep asking for more. Each new connection is a door into other people’s lives: contacts, group chats, shared calendars, the photo album your family keeps together. Give it only what the task in front of you needs. And leave out the threads that carry someone else’s secrets: a friend’s medical news, a colleague’s job search, a family member’s divorce.

**6. Remember that encryption stops at the other person’s screen.** End-to-end encryption protects a message on its way to your friend. It does nothing once their agent reads it on their phone. For a conversation that would hurt someone if it leaked, assume the person on the other end might have an agent reading along. Use disappearing messages, or have that conversation by phone.

**7. Ask, and tell.** Most of us have learned to ask before posting a photo of someone else’s kid. This is the same courtesy. If you use an agent that reads your messages, tell the people you talk to most. If you’re about to share something sensitive, it’s fair to ask whether anyone else is reading. It will feel awkward at first. So did asking about photos.

## And If You Work With Client Data

Anyone who handles other companies’ information has a bigger version of this problem.

**8. Keep work accounts off personal agents — and know what’s coming in the side door.** When you connect your work email to a consumer agent, client proposals, pricing and contracts move into a system your clients never approved. Plenty of client agreements and NDAs don’t allow that. And agents are also coming in the side door: Dots already works inside Slack and Teams. Others are drawing their own lines. Amazon [blocked Muse from shopping on its site](https://www.geekwire.com/2026/amazon-blocks-metas-muse-ai-assistant-in-new-standoff-over-agentic-shopping/), saying it never identified itself and appeared to store customers’ login credentials. [Charles Schwab is blocking Muse from moving client money](https://gizmodo.com/after-muse-launch-u-s-banks-are-saying-hell-no-to-ai-agents-moving-money-2000827221), and its chief executive called agents that act without human approval “a leap too far.” Your agency deserves the same clarity, written down in a policy your team has actually seen.

**9. Ask your tools what your clients ask you.** Before one agency would start a pilot with Signal78, it sent us a due-diligence questionnaire. Do you train on our data? What do you remember? Will you delete our material if we ask? Who can see it? They’re the right questions to ask any agent before it gets near your client’s work. If a tool can’t answer them in plain language, it hasn’t earned the access.

## Being the Human in Someone Else’s Loop

Last year, I wrote that responsibility for our data, and everyone else’s, “belongs squarely with us now.” That’s still where it sits today. I’m less sure it’s where it belongs.

The president of the Signal Foundation, the nonprofit behind the encrypted messaging app Signal, calls the idea that Muse users are making an informed choice “laughable.” She [argues that the burden needs to move](https://www.theverge.com/tech/1011938/muse-ai-data-friends-consent) “from the person who clicked on a consent prompt to the companies that built these underlying systems.” One privacy law professor wants AI agents treated as an ultra-hazardous product, with strict liability. I hope they get there.

Until they do, keeping a human in the loop means more than watching what your own agent does. It means remembering that your inbox is full of people who trusted you, not Meta, not OpenAI and not whichever agent comes next.

You can agree to an agent’s terms for yourself. You can’t agree for your friends.

![Matt Cyr](https://signal78.ai/hubfs/signal78-site/headshot-matt-cyr.jpg)

Founder & President

### Matt Cyr

Nearly 30 years leading marketing teams, from Boston’s hospitals and universities to eight years running strategy and new business at a 100-person agency. Founded Loop in 2024.

## More insights.

[See all insights→](https://signal78.ai/insights)

### [![](https://substack-post-media.s3.amazonaws.com/public/images/ae75c748-4304-4dcd-afe4-39aa21d658bf_1897x1325.jpeg) AI AgentsOctober 13, 2026 Cute AI Agents Are Sweeping Something Under the Rug Read the post→](https://signal78.ai/insights/cute-agents-are-sweeping-something-under-the-rug)

### [![](https://signal78.ai/hubfs/Gemini_Generated_Image_tlr2qftlr2qftlr2.jpeg) AI StrategyOctober 6, 2026 It’s Q4 — How’s Your AI Plan Coming? Read the post→](https://signal78.ai/insights/its-q4-hows-your-ai-plan-coming)

### [![A polybit figure pulls a big lever and every lamp on a console glows teal, with Siggy beside him.](https://signal78.ai/hubfs/insights-headers/insights-header-signal78-live.jpg) Signal78September 28, 2026 Signal78 Is Live. Here's What It's Proven — and What's Next. Read the post→](https://signal78.ai/insights/notes-from-the-build-where-signal78-stands-this-fall)

Every week

## Stay in the Loop.

Get next week’s issue in your inbox: what changed in AI, and what it means for your agency.

![Signal78](https://signal78.ai/hubfs/signal78-site/logo-signal78.png)

The agentic operating platform for agencies. It runs the routine and signals your people when their judgment matters.

[Home](https://signal78.ai/) [Platform](https://signal78.ai/platform) [Services](https://signal78.ai/services) [About](https://signal78.ai/about) [Insights](https://signal78.ai/insights)

<https://www.youtube.com/@signal78-ai> <https://www.linkedin.com/company/signal78/>

© 2026 Signal78 [Privacy policy](https://signal78.ai/privacy)

Cookie settings

×

## Book a 30-minute demo

Pick a time that works. You'll see Signal78 running on agency work like yours.

×

## Talk to us

Tell us a little about your agency and what you want AI to change. We'll reply within one business day.

×

## Can’t we just build this in Claude?

The white paper on what it takes to keep a homegrown AI system current, secure and separated by client.

```json
{
  "@context" : "https://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Matt Cyr",
    "url" : "https://signal78.ai/insights/author/matt-cyr"
  },
  "dateModified" : "2026-10-21T16:45:08.565Z",
  "datePublished" : "2026-10-21T14:21:17.000Z",
  "headline" : "You Can Say Yes to Muse – Just Not on Behalf of Your Friends",
  "image" : [ "https://signal78.ai/hubfs/2026-10-21_asset_blog-friendsdata-robot_r01a.jpg" ],
  "mainEntityOfPage" : {
    "@id" : "https://signal78.ai/insights/you-can-say-yes-to-muse.-your-friends-cant",
    "@type" : "WebPage"
  },
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "url" : "https://signal78.ai/hubfs/Loop%20Logos/Loop_Primary.svg"
    },
    "name" : "Signal78"
  }
}
```