Is the Fed ready for an AI economy?

But the Fed may be ignoring another AI issue that could turn out to dwarf all others: AI’s coming ability to create money and credit—and thereby determine the money supply in our economy. This may sound like science-fiction, but AI will likely begin to control the extension of credit, and in effect the money supply, in the next few years. That’s a major issue for the Fed if it can’t easily observe the manner in which AI makes its decisions or constrain it to follow the central bank’s underlying interest-rate policies. Here’s how it will work: Individual users will start with personal-assistant AIs, as many of us already have. As guardrails are established on AI’s behavior and their actions refined, users will employ them as purchasing agents, giving them the authority to use credit on behalf of the user. On the banking side, business operations on the decision to extend credit or not will start to be tasked to AIs. These AIs will assess creditworthiness and make an assessment to create a loan or authorize a credit purchase. The aggregation of millions of such decisions will make AI a major influence on the expansion or contraction of the money supply. Most likely this will first take place in the “shadow banking" part of the economy—lightly regulated fintech companies that are experimenting with new forms of credit creation—which is outside the direct purview of the Federal Reserve. The Genius Act, which passed the Senate last week, makes these events even easier to envision. The act would formalize the stablecoin industry. Stablecoin is a Treasury-backed cryptocoin that allows transactions to move onto the blockchain, a decentralized digital ledger outside the traditional banking sector. Companies like Amazon and Walmart have thrown their hats into the ring to create their own stablecoins, skirting the traditional banking system in all transactions. With these elements in place, the Federal Reserve will face the problem of an AI purchasing agent talking to an AI credit agent with the only constraint on them being the underlying code dictating their behavior and the instructions users give them. Further, these two agents may conduct transactions using a form of currency that is lightly regulated and doesn’t move through the traditional banking system. Since these interactions would take place in the shadow banking sector, the Fed won’t see first movements in where the money supply is expanding or contracting, one of the main things central bankers need to observe to do their job correctly and guide the economy. And as AIs make financial decisions, how can anyone be certain that they will respond to Fed policy in the same way as our historical data tells us humans respond to Fed policy? There are possible private-sector solutions to mitigate some of this, like insurance issued on the behavior of AIs making decisions. But if the Fed doesn’t have a clear picture of money creation by AIs or a direct way to intercede in AIs’ ability to change the money supply, it is in for a world of problems in an AI-driven economy. The Fed needs to address this sooner rather than later. If the Genius Act becomes law, it will only make the problem more urgent. The pace at which personal AI products are being rolled out signals that this reality is fast approaching. If the Fed doesn’t have clarity on these innovative areas of the shadow banking sector and a playbook to deal with AI money creation, then it will realize the depth of this problem far too late to do anything about it. Mr. Horstmeyer is a professor of finance at George Mason University.