A Chainlink co-founder said on stage that if his AI ever suggested a $450 pair of running shoes, he’d shut the whole thing off that day. Everyone’s talking about AI that can spend money. Buy stuff, move cash, pay bills. Nobody’s asking the boring question: who decides how much the AI is allowed to spend? That was the real story of this panel. Here’s what I took from it. 1. The old fight is over Three years ago the debate was whether banks would ever touch blockchains. @SergeyNazarov pointed out that it got settled while most people weren’t looking. Swift, DTCC and the big clearing houses all run their own versions now. So the question moved. It’s not “will banks use this.” It’s “who’s the customer?” 2. Versions of the future Version one is easy to picture. AI is your hands. A back-office person says “move this collateral,” and the AI does it. Version two is the weird one. The AI becomes its own kind of customer. It makes buying and money decisions with way less human input than we’re used to. He was clear this isn’t happening yet. But at the speed AI is moving, he thinks it’s possible. And if the buyer is a machine, a lot of what companies pour money into stops mattering. Faster page loads. Nicer buttons. A cleaner checkout. A machine doesn’t click buttons. His joke: you know those screenshots of a bank website with startups circling each section they’re replacing? Now the whole page is one arrow pointing at “AI.” Whoever the AI picks wins. Anyone still polishing buttons for humans is playing the old game. 3. The permission slip If an AI has a bank account, what stops it from going wild? His answer is a public registry. Every AI agent gets an ID. Next to that ID sits a file listing what it’s allowed to do. Something like: “this agent can spend up to $100,000 on Mastercard’s agent payment system.” If it asks for $1 million, the system says no, then pings the human. Banks and payment companies check the registry before the AI moves a cent. No entry, no money. Why doesn’t this exist already? Because everyone has to be able to read it and write to it. Banks, AI companies, regular users. Nobody wants one company owning that. He pointed to the LEI, the existing ID system for legal entities, as a promising base. They recently published a report on it, and @chainlink is connecting its systems to theirs. 4. So what’s left for humans? Depends on how good the AI is. His example was shoes again. He usually buys his at 50% off. If the AI finds them at 60% off, why would he ever step in? You treat it like your best employee. You give them more work because they’ve earned it. Then he admitted he might turn into a drooling idiot who can’t do anything, and hoped he’d use the free time to read more books. 5. Where the money goes Elon Musk said money will look totally different by 2036. Sergey thinks that’s a fair date. His reasoning is simple. Digital versions of cash, stocks and funds are growing into a second system next to the old one. The old rails are rigid and won’t change much. The new one only gets better. AI takes the path of least resistance, so it’ll walk down the new one. Behind the scenes, banks are making digital versions of their deposits. Digital dollars like @tether and @circle keep growing. He said Tether already gives people in emerging markets a way to hold dollars without a bank account, and he put the value at tens of billions of dollars. Central banks don’t agree with each other. Europe and the Middle East lean in. The US says no. The messy part is getting all of these to swap with each other. A weak bank’s digital deposit isn’t the same risk as a strong bank’s, or a central bank’s, or a fintech’s. Nobody has fully figured out how to price that. He said that’s what Chainlink is working on now. ethereum:0x514910771af9ca656af840dff83e8264ecf986ca

Korean institutions want tokenized deposits, stablecoins and RWA. The market is there. Per Chainlink’s announcement it’s about $68B, and the won ranks second in crypto trading volume, behind only the dollar. But a bank can’t just plug into a public chain. It needs data feeds, cross-chain messaging, wallets, transaction tracking, compliance. Nobody at a Korean bank wants to assemble that from crypto parts. Why nobody solved it Regulation, mostly. And a turf war. Korea’s Digital Asset Basic Act has slipped repeatedly. The FSC and the Bank of Korea disagree on who controls stablecoin issuance, including whether issuers should be bank-led consortiums. A unified bill was expected in September, with the target still being passage before the end of 2026 Tokenized securities moved first. Amendments to the Electronic Securities Act and the capital markets law passed in January 2026. So the order in Korea is probably securities tokenization first, won stablecoins later. What was built Two separate things, and people are mixing them up. First, Project Hangang. It’s the Bank of Korea’s deposit token pilot, and LG CNS is the lead contractor. Phase 1 started in April 2025, ran about three months, onboarded up to 100,000 people and processed 118,000 test payments. Phase 2 expands to nine banks and tests real government subsidy payments. Second, KITL. This is LG CNS’s own platform for stablecoins, tokenized securities and RWA. It provides wallets, transaction processing, gas fee payment on behalf of users, and onchain transaction data collection. @circle, @Securitize, @CantonNetwork, @chainalysis and @chainlink are the named partners. Chainlink’s slot is the second one. The announcement says it will plug cross-chain, data and orchestration into LG CNS’s platform. I found nothing saying Chainlink is inside Hangang itself. Hangang is deposit tokens on central bank infrastructure, which is a different design. How this could reach ethereum:0x514910771af9ca656af840dff83e8264ecf986ca Launched in August 2025, it uses Payment Abstraction to convert revenue into ethereum:0x514910771af9ca656af840dff83e8264ecf986ca . Enterprise clients pay in fiat or stablecoins, the system swaps it on DEXs, and the ethereum:0x514910771af9ca656af840dff83e8264ecf986ca goes into the reserve. Chainlink says its enterprise deals have already generated hundreds of millions of dollars in revenue, and it doesn’t plan withdrawals for several years. So the pipe exists: LG CNS pays for services → revenue converts to ethereum:0x514910771af9ca656af840dff83e8264ecf986ca → reserve grows. Crypto has one narrative left, and it’s RWA. In that world the chain matters less and the plumbing matters more: oracles, proof of reserve, cross-chain messaging. That’s Chainlink’s strongest position. But plumbing only works as a thesis if usage turns into token demand. The reserve is the first real attempt to wire that up.