I was on the earnings call when Mark Zuckerberg described Muse as a system that should be everywhere. You can almost feel the scale in his words: billions of agents, each answering someone’s question. I want to walk you through what that ambition will cost us, and what it might cost you.
On the call, Meta said it expects Muse to serve billions — the pitch and the pull
Meta’s goal is blunt and audacious: “We expect to grow Muse to serve billions of people,” Zuckerberg said. The language is both mission statement and sales strategy: get millions of people talking to Muse, then let those Muses talk to each other.
I’ve watched tech pivots before, and this one has the familiar momentum of a company steering itself toward a new center of gravity. Threads outpacing X in monthly active users and 10 million people in the U.S. and Canada already setting up personal Muse agents give that gravity real mass.
One image that sticks is a cathedral of servers—vast, humming, built to hold millions of small transactions every second.
Meta says Muse’s most common use case right now is shopping: more than a quarter of users buy through it, with an average purchase of over $200 (€186). That’s an easy hook for broader adoption: convenience that pays for itself.
What is Meta’s Muse and how does it work?
Muse is a personal AI agent inside Meta’s apps that can chat, search, shop, draft, and coordinate with other agents. It pulls on your account activity, your messages, and external web data to answer queries and complete tasks. Meta’s argument: millions of Muses acting together will be more useful than isolated assistants.
A New Mexico ruling is a reminder — privacy errors have consequences
A judge in New Mexico ordered Meta to pay $25 billion (€23.3 billion) for misleading users about privacy tied to Cambridge Analytica.
If you track Meta’s legal history, that verdict lands like a pebble thrown at a glass house. Muse’s rollout hasn’t been free of stumbles: reports that agents accessed private texts, a user’s home address allegedly given to a stranger, researchers saying Muse builds exhaustive, updated files on people in a user’s life—and Amazon temporarily blocking Muse from shopping on its platform over credential concerns.
Those incidents aren’t theoretical. They map to real-life harms: leaked addresses, unauthorized account actions, and the frightening possibility that an agent could betray the trust you place in it.
Will Meta share my data if Muses communicate?
Short answer: Meta says no, but the practical answer is messier. When agents collaborate across accounts and services, the boundaries around what data moves where blur. Given recent reports and the company’s past, you should expect aggressive data engineering and legal arguments over what counts as shared data.
On the spreadsheet, the math is plain — Muse is expensive to scale
Meta told investors it expects to spend at least $135 billion (€125.6 billion) this year, mostly on AI infrastructure.
Running billions of agents is compute-intensive now; scaling will multiply that cost. Meta’s cash flow is under pressure: after an $8 billion (€7.4 billion) hit in a previous quarter, its free cash flow turned negative—negative $2.12 billion (€2.0 billion) in the most recent report.
To raise money without diluting control, Meta plans to hand existing shareholders a new class of non-voting capital stock in a two-for-one split next year. That’s a financing move with governance implications: it buys cash room while preserving voting power at the top.
How much will Muse cost Meta to run?
Expect continual, large-scale capital spending: more data centers, specialized chips, and network capacity. Meta’s current guidance—$135 billion (€125.6 billion) this year—is a sign that Muse is not a small operational line item, it’s the company’s largest bet.
In boardroom terms, control rules — and the politics are personal
Meta introduced a Founder Alignment Agreement that keeps Zuckerberg in California and lets him sell non-voting shares if a proposed 5% billionaire wealth tax (Proposition 40) passes.
You can read this two ways: as a practical hedge for a billionaire, and as a power-preserving maneuver that shapes how Meta will be funded and governed as Muse expands. The split gives investors a new instrument, and gives Zuckerberg a path to liquidate assets without surrendering control.
That mix of capital engineering and political positioning affects product decisions. When a CEO can monetize growth without losing control, risk thresholds for privacy trade-offs and aggressive product rollout shift.
Signals and hazards — where I would watch closely
My attention is on three things you should watch: data governance, inter-agent protocols, and third-party integrations.
Researchers say Muse creates detailed dossiers on interpersonal networks. Amazon flagged credential handling. There was a zero-day that allowed account hijacking—patched, but revealing where attackers can probe.
The second image I’ll leave you with is an orchestra of bots: if the conductor (policy, law, or platform rules) falters, discord follows quickly.
If you care about privacy, you should ask how agents exchange information, who can audit those exchanges, and what real-world remedies exist when things go sideways.
I’ve been watching this sector long enough to know that scale rewrites the rules. You’ll be offered convenience that feels inevitable; the trade-offs will be baked into the plumbing, not the interface. Which trade will you accept, and which will you refuse?
Meta’s Muse is a bet on scale, capital, and trust; the outcome will hinge on the rules we demand and the controls we hold—so what rules will you push for?