AI in action · Tools

Meta's Muse Agent Says It Can Manage Your Money. An Identity-Theft Charity Says "the Technology Is Too New" to Recommend

Meta's Muse agent offers to track spending, negotiate bills and buy things, and more than 10 million people in the US and Canada have set it up. Yahoo Finance reviewed those money features on October 22, 2026. Meta says Muse asks for approval before spending and uses single-use card numbers, but also that it "isn't immune to attack" and "will sometimes make mistakes." The Identity Theft Resource Center does not recommend it yet.

Oct 24, 2026

AI agents are moving from answering questions about money to handling it, and Meta's Muse is the largest test so far. More than 10 million people in the United States and Canada have set it up since its September launch, according to chief executive Mark Zuckerberg, and Meta said on its earnings call that it expects more than a billion people to use the service eventually. Its chief AI officer, Alexandr Wang, promoted the money features at the company's Connect conference with a #MuseMoneyChallenge: let the agent find $1,000 of savings. Yahoo Finance's Kendall Little looked at what that involves in a review published on October 22, 2026.

The list is long. Muse can track spending and set budgets, audit subscriptions and flag unexpected charges, find duplicate charges, negotiate bills, including by calling providers, search unclaimed-property databases and file claims, and find deals and track prices. Meta describes it as an agent that, once a person shares a goal, "helps them develop a personalized plan and coordinate their time and resources, then advances the work on its own. It can open a browser, fill out forms, and negotiate on their behalf." Purchases happen only after the user approves them, and where no card is stored the agent uses single-use card numbers. Meta says it stores passwords and payment methods so that they are hidden even from the agent, and keeps a trail of past and planned actions.

Meta's own caveats are as direct as any critic's. The company says Muse "isn't immune to attack" and "will sometimes make mistakes." Eva Velasquez, chief executive of the Identity Theft Resource Center, a US non-profit, was blunter: "Right now, the technology is too new. The data sharing implications are still unknown … It's just not something we're going to recommend." The review's suggestion for privacy-minded users is a halfway house: upload a statement with personal details removed and let the agent analyse it, rather than connecting the account.

The timing adds context the review could not. The same week, Axios reported that major US banks are finding unvetted third-party agents inside customer accounts, often logged in with credentials customers handed over, which most account agreements forbid, and JPMorgan and Charles Schwab have blocked outside agents altogether. An agent that negotiates a cable bill on the phone is one thing; an agent with a bank password is another, and the liability when it errs sits, for now, mostly with the customer.

The useful distinction is between reading and acting. Letting an agent read a redacted statement and suggest which subscriptions to cancel carries little risk and can save real money. Letting it move money, sign up for services or speak for you on a call is a delegation of authority, and the honest position of both the company and its critics is that nobody yet knows how often it will go wrong.

Sources: Yahoo Finance (Kendall Little) — Meta's Muse says it can manage your money. Should you let it? (Oct 22, 2026); Axios — Scoop: Unvetted AI agents are accessing U.S. bank accounts (Oct 22, 2026)