Meta's Muse is a data collector in disguise

Meta's new AI agent is impressive at browsing the web, but it's really designed to harvest your personal information.

By ·

Sep 20, 2026

14 cards · 4m · tap to begin

From · · · 4m

Meta's Muse is a data collector in disguise

Meta's new AI agent is impressive at browsing the web, but it's really designed to harvest your personal information.

In brief

Meta's new AI agent is impressive at browsing the web, but it's really designed to harvest your personal information. AI agents like Muse prioritize data collection over user benefit, automatically opt users into training their models, and risk making people passive and dependent on machines to make decisions for…

Meta launches Muse AI agent

Meta's new app Muse helps automate everyday tasks like booking reservations and managing your inbox. It downloaded over 900,000 times in its first week and works through WhatsApp and Instagram.

“Hand off everyday tasks like finding deals, booking reservations, and managing your inbox.”

Muse browses the web impressively

Muse browses the web impressively

Muse uses a virtual machine to browse the internet on your behalf, searching and clicking through websites with remarkable accuracy. When asked to order breakfast from a local bakery, it successfully navigated the site and added items to the cart.

“Muse rarely seems to get lost while browsing.”

Facebook Marketplace integration shines

Muse impressed most when searching Facebook Marketplace for local couches within specified parameters and offering to message sellers. It even followed up the next day to remind the user about their favorite option.

Muse stores your preferences permanently

Muse creates a Memory document that stores details about your interactions and preferences. While you can manually edit or request deletion, Meta doesn't offer a toggle to turn off memory collection entirely.

“Your curated long-term memory: durable facts, preferences, and commitments.”

Constant requests for more data

Muse repeatedly suggests connecting more information sources like checking and savings accounts, email inboxes, and photos of documents. Each feature offer is framed as personalization but requires sharing more personal data.

Users automatically opted into AI training

Muse users are automatically opted in to having their interactions used for AI model training. Meta says the data is sanitized, but the process is unclear and may include context from connected data sources like email and bank accounts.

Meta defends data collection as necessary

Meta argues that collecting user data is critical for training better models and improving performance. The company plans to release confidential versions later that cryptographically prevent Meta from accessing the data inside.

“Every Muse user gets a better personal agent as we all collectively use the product.”

Privacy advocates call opt-in manipulative

Consumer advocates say Meta's default opt-in for AI training shows the company hasn't learned from past mistakes. They point to Meta recently forcing Instagram users into an AI deepfake tool as evidence of manipulative design patterns.

“This tells me they have not learned from past mistakes.”

Agent browsing influences your ads

While Muse data isn't directly shared with advertisers, the agent's browsing activity can influence what ads you see on Instagram through web-tracking algorithms. Making a dinner reservation or shopping could indirectly shape your ad experience.

Sponsorship deals could bias recommendations

Consumer advocates worry that companies could pay to have their products prioritized in an agent's output. Meta's Chief AI Officer hinted at exploring non-ad revenue opportunities in Muse without disclosing specifics.

Automation strips away human joy

Letting an agent handle all your decisions removes the pleasure of browsing, exploring, and discovering your own preferences. The inefficiency of shopping is actually where people build taste and figure out what they genuinely enjoy.

“That's kind of the joy of being human, getting to test things out and figure out what you like.”

Agents cause cognitive degradation

Researchers warn that AI agents make users more passive and dependent on external tools. People may become less able to make independent decisions and assess what the agent is doing on their behalf.

“The design of AI agents is such that it disengages users.”

Personalization pulls you deeper in

As Muse learns more about you, it becomes more personalized and trustworthy, which paradoxically makes you more likely to share even more private information. The agent mirrors your speaking style as it collects more data.

One user deleted Muse after testing

Image may contain Adult Person Face Head Photography Portrait Art Smoke Pipe and Drawing

After experimenting with Muse, the reviewer decided the app was more focused on data collection than actually helping users. Even after deleting it, Muse sent a final notification asking to connect more apps.

“Connect your apps so I can do more.”

What to remember

AI agents like Muse prioritize data collection over user benefit, automatically opt users into training their models, and risk making people passive and dependent on machines to make decisions for them.

Read the original on WIRED

React

Sign in to react and comment.

Comments (0)

Life is short. Keep it sweet. Respect others' opinions and be kind!

    Recommended next

    More decks on meta and related topics.