Muse wants to manage personal finances, but Meta must sell trust first

Meta introduced Muse as a personal agent for mainstream consumers. It runs in a dedicated virtual machine, uses connected services and, with user permission, can send emails, make purchases or manage other tasks. Meta says it designed the product for billions of people, although the launch is currently limited to the US on iOS, Android and the web.

In a TechCrunch discussion, reporter Sean O'Kane described an early experience: Muse suggested searching for unclaimed funds and found money that should now arrive by check. He considered the result useful but one-off. Further scenarios, such as finding unused subscriptions or duplicate charges, would require access to credit cards, Gmail and other sensitive services.

Muse targets consumers while rivals pursue companies

The strategy fits Meta's established strength: Facebook, Instagram and WhatsApp already occupy a place in the daily lives of a vast audience. A personal agent does not need to win over corporate procurement first. It can attach itself to accounts and communication habits people already have.

That changes the measure of success. Finding money once attracts attention, but recurring value requires small useful interventions every week. Muse needs to be helpful enough that users gradually connect calendars, email, finances and social networks.

The most valuable features require the data people guard most closely

Meta says conversations and data inside the virtual machine are not shared with its advertising systems. Muse's web activity can still influence ads indirectly, just like an ordinary website visit. The option allowing interactions to be used for model training is enabled when a person first uses Muse, although it can be switched off retroactively.

The company also describes Sentinel, a separate permission layer, and approvals for sensitive actions. A Confidential VM intended to cryptographically prevent Meta from accessing its contents is planned for later this year. Trust therefore rests on a mix of current controls and a future promise.

One early win does not establish recurring utility

Finding money is a compelling demonstration, but the same value cannot recur every week. The TechCrunch account captures the gap between an impressive first day and a reason to leave long-term account access in place.

One test cannot establish adoption or the service's overall safety. It does expose the central product tension: the more useful Muse becomes, the broader the permissions it needs.

Only recurring tasks will keep accounts connected

Repeatable use cases and the rate at which people disconnect services after the first week will matter. Public security audits of the Confidential VM and a clear record of which data enters training or can indirectly affect advertising will matter too.

Muse succeeds if it saves measurable time or money each month without surprising interventions. Integration with many accounts is a distribution advantage, but it is also the highest barrier to connecting them.

Lilith's verdict

Muse found its user a check, but its real cash register is permission to access email, cards and calendars. Meta must show that the cashier is not reading the shopping list for someone else.