Muse Found Real Unclaimed Cash, But Testers Still Don't Trust MetaMuse Found Real Unclaimed Cash, But Testers Still Don't Trust MetaMuse Found Real Unclaimed Cash, But Testers Still Don't Trust MetaMuse Found Real Unclaimed Cash, But Testers Still Don't Trust Meta
September 28, 2026
Meta's new personal AI agent (software designed to act on a user's behalf rather than simply answer questions), Muse, found real unclaimed money sitting under Sean O'Kane's name within weeks of its release, a working financial result from the product Meta positioned as the
Meta's new personal AI agent (software designed to act on a user's behalf rather than simply answer questions), Muse, found real unclaimed money sitting under Sean O'Kane's name within weeks of its release, a working financial result from the product Meta positioned as the centerpiece of Meta Connect. But the more consequential finding in TechCrunch's own hands-on testing sits beneath that headline: the same tester who got a real check coming in the mail says he still does not trust the company that built the tool that found it. Mark Zuckerberg, Meta's CEO, told the Connect audience the company plans to "push AI features everywhere," yet adoption depends on a trust his own testers say Meta has not earned. That gap, a feature that worked paired with trust that did not follow, is the question hanging over every large platform trying to turn a chatbot into an agent with access to someone's bank account.
What Happened
Muse debuted as the centerpiece of Meta Connect, Meta's annual product event, where Zuckerberg said the company intends to "push AI features everywhere" across its apps and devices. Meta positioned the product as a consumer-facing counterweight to more enterprise-focused announcements arriving from rivals the same week.
TechCrunch weekend editor Anthony Ha published an account of the product on September 27, 2026, at 12:57 PM PDT, built around firsthand testing by Equity podcast co-host Sean O'Kane, who had been using Muse for what he described as "a couple weeks," alongside co-host Kirsten Korosec. On first download, Muse did not immediately pull in O'Kane's Threads, Instagram, or Facebook data or context, treating him, in his words, like a stranger. That restraint, O'Kane said, made him more willing to trust the app early on. Muse also surfaced proactive suggestions, including a prompt to scan for unclaimed funds; O'Kane followed it and found real unclaimed property money registered in his name, with a check now coming by mail. He noted, however, that Muse's behavior shifted with continued use, actively trying to pull in his social data over time so it could "learn all this stuff about" him.
Why It Matters
The unclaimed-funds result is a genuine, verifiable outcome, not a hypothetical use case. Yet O'Kane's verdict on the broader product was blunt: "I don't know that I will ever trust Meta the same way." That statement, from someone who tested Muse for weeks and personally benefited from it, signals that a single working feature does not automatically transfer into confidence in the company behind it.
O'Kane grounded his skepticism in Meta's business model rather than any specific technical flaw in Muse: "Meta's business is to sell you ads. And yes, they'll make the argument that the more they know about you, the more accurate and interesting the ads will be, wake me up when we get to that fever dream." That framing treats Muse's data collection not as a feature-development question but as a monetization question, since every additional signal Muse gathers about a user is also, under Meta's existing business, an input to ad targeting.
O'Kane also downgraded the unclaimed-money win itself, calling it "a party trick-type thing," meaning a novelty that works once and does not obviously justify daily use. Korosec pushed back directly on air: "I mean, you just killed your own argument. I don't see how that wouldn't become wildly popular." That exchange captures the article's central tension in miniature: the same feature can read as a one-off gimmick or as proof of concept for a much larger financial-agent product, and which reading wins depends less on the feature than on whether users trust the company running it.
Competitive Landscape
Muse launched into a week when the AI industry's biggest players moved in the opposite direction. OpenAI and Anthropic both shipped new model updates the same week, both aimed at enterprise and coding use cases rather than consumer-facing personal agents. Meta's decision to build a consumer product instead reflects the company's established strength embedding tools into everyday use through Facebook, Instagram, and WhatsApp, rather than competing head-on in the enterprise lane.
O'Kane placed Muse directly against a named predecessor: "this is roughly Meta's kind ground-up version of an on-your iPhone, or on your Android, app of OpenClaw, which we talked about a couple months ago, which Meta went out and basically bought and integrated those folks' work."
Three existing reference points frame how far Muse still has to go:
- Rocket Money: an existing service O'Kane cited that already performs subscription-cancellation and double-charge detection, the exact category Meta has talked up expanding Muse into.
- Apple's Siri: O'Kane said he trusts the new Siri on his upgraded iPhone with sensitive financial and personal data more than Muse, specifically because Apple's business is not ad-based.
- OpenClaw: the product O'Kane says Meta effectively rebuilt and integrated into Muse's foundation, rather than a from-scratch Meta effort.
Independent analyst commentary specifically on this announcement was not publicly available at publication time.
The Bigger Picture
The structural problem Muse faces is not a bug that a future update fixes. A personal AI agent needs deep access, financial accounts, email, browsing history, to do the tasks Meta wants it to do, and Meta's core revenue line remains advertising built on exactly that kind of behavioral data. Every improvement that makes Muse more useful as a financial agent also makes it a richer source of ad-targeting signal, and O'Kane's own testing suggests users notice that tension even when a specific feature works as promised.
That is why the specific wording of O'Kane's verdict matters: "I don't know that I will ever trust Meta the same way." The phrase "the same way" implies a prior baseline of trust that has already eroded, independent of whether Muse itself misbehaves. A company can ship a technically competent product and still inherit skepticism accumulated from its existing business practices. Meta's design choice not to pull in Threads, Instagram, or Facebook data on first use appears to have been a deliberate attempt to manage that skepticism, easing users in before asking for deeper access. Whether that sequencing holds as Muse reportedly moves toward deeper financial-account access is a question the current reporting does not resolve.
What's Next
Meta is reportedly moving Muse toward the subscription-cancellation and double-charge detection category that Rocket Money already serves, extending the unclaimed-funds trick into a broader financial-agent role. That would put Muse in direct competition with Rocket Money's existing service, but with the added reach of Meta's consumer install base across Facebook, Instagram, and WhatsApp.
The open question is whether Meta can convert a one-time "party trick" into a feature people use repeatedly, the same question Korosec raised in pushing back on O'Kane's dismissal. Meta has not publicly responded to the trust concerns O'Kane raised.
The most telling detail in TechCrunch's testing is not the check coming in the mail, it's the sequencing: Muse earned early trust precisely by acting like it knew nothing about its user, then spent the following weeks trying to undo that restraint. Meta built a product that works. Whether it can build a product people believe in is a separate contest, and on the evidence of its own testers, Meta has not won it yet.
For a household juggling recurring charges across streaming, fitness, and cloud storage subscriptions, a financial agent that scans linked accounts for duplicate charges and forgotten renewals is a genuinely useful pitch, the same one Rocket Money already sells today. The difference with Muse is who is doing the scanning: a company whose primary business is advertising, offering to do the kind of account-linked scanning Rocket Money already does. Whether that trade is worth it will likely come down to whether Meta can prove, feature by feature, that the data it collects to cancel a subscription does not also end up shaping the ads a household sees.
-- Aria Lin, Enterprise Technology Analyst
Sources: TechCrunch · Meta Connect keynote remarks