NFTs were written off as a joke, then quietly became real infrastructure. AI in credit unions is following the same arc, and the industry is reading it wrong.
In October 2021, a cartoon ape sold for $3.4 million. It was a JPEG, one of the Bored Ape Yacht Club images, auctioned at Sotheby's. Anyone could right-click and save the same file for nothing. Celebrities adopted the apes as profile pictures. Within a year the market had collapsed, and the apes became the reference point for everything hollow about the crypto era, proof for most people that NFTs had been a con from the start.
That verdict was fair to the apes. It was not fair to the technology underneath them.
The apes were never the technology
While the speculators chased cartoons, four of the largest luxury groups in the world were building on the same infrastructure. In 2021, LVMH, Prada, Richemont (the owner of Cartier), and the parent of Diesel and Maison Margiela founded the Aura Blockchain Consortium. There was no marketing campaign and no celebrity partnership. The problem they were solving was unglamorous and expensive: counterfeiting, which costs the industry tens of billions of dollars a year and had defeated every hologram and serial number ever printed. Their answer was to assign each authentic product a non-fungible token that stays with the item permanently and records its ownership history. It is the same core technology as the ape. By the consortium's own count, more than 50 million products now carry one. When a customer buys a Louis Vuitton diamond, its certificate of authenticity is, in technical terms, an NFT.
One tool, two uses. One group traded cartoons at absurd prices. The other made a counterfeit handbag nearly impossible to pass off. Only the first became a story.
Credit unions are positioned to make the identical mistake with artificial intelligence.
The problem is the word "slop." Slop is conspicuous because it is cheap to produce. A fraudulent listing, a malformed AI image, a chatbot that cannot locate an account: none of it takes much effort, and all of it circulates easily. The useful applications are quieter by nature. No one forwards a screenshot of a model that flagged a compromised card before it cleared. So an entire technology gets judged by its most visible failures. That is how the apes were misread, and it is how AI is being misread now.
What AI in credit unions already does
Set the demonstrations aside and consider what AI in credit unions already does, not in a pilot but in daily operation. AI models screen card and account activity for fraud in real time. They support loan underwriting, pulling signal from thin credit files that traditional scores miss. They read and verify the documents that accumulate in a lending file. They watch for the transaction patterns that anti-money-laundering rules require someone to catch. They handle the first layer of member questions so staff can take the harder ones. None of it is speculative, and much of it arrived through core and vendor systems that institutions never chose deliberately. In January 2026, the NCUA published an updated AI resource hub. The regulator has moved past the question of whether.
The hype is slop, too
The skepticism has to run both directions, because overstating what AI can do is its own form of slop. Fully autonomous systems that run a complete workflow without human oversight are not suitable for high-stakes functions inside a financial institution, and will not be for some time. Any vendor offering a platform that underwrites, discloses, and books a loan with no person in the loop is presenting a demonstration, not a deployment. Adverse-action requirements alone will keep a human in that process for years. The functional applications are real. The autonomous ones still have to prove themselves.
None of this requires a credit union to be an early adopter. It requires the discipline to separate the excess from the substance beneath it, which is harder than either enthusiasm or dismissal. The institution that dismissed NFTs as speculative nonsense felt vindicated by the crash and never registered that the same technology now authenticates more than 50 million luxury goods. The institution preparing to dismiss AI on the strength of its slop is in the same position, one technology later, with considerably more of its own capital at stake.