Is Meta's Muse the Democrati­sation of AI or a Wolf in Sheep's Clothing?

Muse is a personal AI agent released in the United States in September 2026 by Meta, the parent company of Facebook. It handles tasks through the user’s email, calendar and bank connection. Meta markets it as a personal AI agent built for everyone. In my view, it is an exercise of power, not the democratisation of AI.

In Finland, nobody hands a stranger the keys to their home, their letterbox and their address book. A bank does not show its customers’ letters to advertisers. An AI agent that gets into the same places must follow the same rules. In the EU, the General Data Protection Regulation and the Digital Markets Act require it.

What is Meta Muse and what does it do?

Meta Muse is Meta’s personal AI agent. According to Meta’s own announcement, it sends emails, books trips, fills in forms and negotiates on the user’s behalf. It keeps working even after the user closes the app, and asks permission before sensitive actions. The user chooses which apps the agent is connected to. Muse has a free basic tier and paid tiers.

Muse operates in the United States on phones, WhatsApp and the web, and on Mac computers. According to TechCrunch, Meta said in its October earnings release that Muse has more than 10 million users and that shopping is one of its most popular uses.

What data does Meta Muse collect, and about whom?

Meta Muse collects data from every service the user connects to it, and also about the people the user talks about. According to a TIME investigation, the agent maintains files on these people and updates them hourly. People who do not use Muse also end up in other users’ agents’ files. Meta’s AI is trained on users’ conversations by default, and the user can opt out.

Muse processes, for example:

• email and calendar
• bank data through the US Plaid bank connection
• files on people mentioned in conversations, messages and emails

The TIME investigation is based on Muse’s internal guidelines. According to it, the files record how people met, what they have quarrelled about, and who has relationships or tensions with whom. The agent also infers goals that the user has not spoken aloud. If you ask Muse to forget something, the original message may not be deleted. According to Meta, agents learn from one another only after names and identifying details have been removed. According to Meta, users’ data is not shared with the advertising system, and users can view and export their Muse files.

Muse keeps files even on people who have never given it permission.

Collected data does not always stay where it is meant to. According to Ars Technica, security researchers found Muse share links in public DNS records and used them to access more than 6,800 conversations. They contained companies’ revenue figures, loan applications and account transactions. Meta has since shut down the sharing feature.

Why is Meta Muse not the democratisation of AI?

Meta Muse is not the democratisation of AI because its business model concentrates power in a single gatekeeper company. The same company’s agent gets access to the user’s email, calendar and bank connection, and the company takes a share of the purchases the agent makes. A product can be available to everyone without power being distributed. Power concentrates with whoever holds the data.

Meta’s chief executive wrote in July 2025 that the company wants to bring personal superintelligence to everyone and put power in people’s hands. According to TechCrunch, Muse’s revenue model rests on Meta keeping a share of the purchases made through the agent. Meta expects Muse eventually to earn more per user than its other apps, such as Facebook and Instagram.

The strongest privacy protection, in which not even Meta could see the content, has been promised later as a separate option, as ABC News also reports. It is therefore not on by default. In the guidelines of its bug bounty programme, Meta itself writes that the Muse agent holds the user’s most sensitive data and can act on the user’s behalf. Facebook, Instagram and WhatsApp are already owned by Meta.

Is Meta Muse coming to Finland, and what does EU regulation require of it?

Meta Muse is not available in the EU, and Meta has not said whether or when it will come to Europe. According to the Spanish outlet Merca2, analysts see EU regulation as a brake on Muse, because the Digital Markets Act (DMA) requires a gatekeeper to obtain the user’s separate consent before combining personal data across services. This is, however, only the analysts’ assessment, not a reason stated by Meta.

The European Commission has designated Meta a gatekeeper under the Digital Markets Act, and its core platform services include Facebook, Instagram, WhatsApp and Messenger. According to the regulation, a gatekeeper may combine personal data between its core platform services and its other services only if the user has explicitly chosen this and given consent. The Merca2 article ties Muse precisely to this rule.

The Irish Data Protection Commission is Meta’s lead supervisory authority under the EU General Data Protection Regulation (GDPR). As of early November 2026, it has not taken a public position on Muse. Neither the European Data Protection Board nor the Finnish Data Protection Ombudsman has taken a position either.

What should Finland do?

Finland must insist in the EU that personal AI agents comply fully with the General Data Protection Regulation and the Digital Markets Act. The rules already exist, and agents need no exemptions from them. Confidentiality of correspondence and banking secrecy are taken for granted in Finland, and an agent that reads your mail and sees your accounts falls within the same circle of trust.

This means three things:

• A gatekeeper company combines data across its services only with separate consent.
• No profiles are built of people who do not use the service.
• The strongest privacy protection is the default for everyone, not a paid extra tier.

Finland is a high-trust society. In my view, a profile, once built, is practically impossible to take apart, so the rules must apply to agents from the start. This is about rules, not about banning American products, because digital independence is not isolation. I have written before about why Finland risks losing control of its own data. For what you can do yourself, I gathered the options in a post on everyday choices for digital independence.

If a free service sounds unbelievably great and advanced, it is worth treating it with healthy suspicion. It is quite possible that you are the product, and that your data is what the provider sells to its other customers.

Real democratisation of AI means that basic-level AI is available to everyone through public services. The library is a good model. It sells nothing, respects privacy and does not collect users’ data for commercial purposes. I am a candidate for the Finnish Parliament (eduskunta) in Uusimaa, and if elected I will carry these principles forward. When AI is offered in libraries, schools and employment services on the same principles, nobody has to buy their privacy back.

Cover photo: “Watching You Watching Me”, Todd Huffman, CC BY 2.0, modified.