Artificial intelligence could eventually predict central-bank decisions more accurately than policymakers themselves, allowing AI agents to exploit or circumvent monetary actions and potentially make markets less informative and more erratic, Princeton University economist Markus Brunnermeier told global central bankers in Jackson Hole, Wyoming. He urged regulators to prepare for what he described as possible large-scale disruption, including by considering simpler and more robust rules. In extreme scenarios, central banks might face pressure to communicate less predictably, hold larger balance sheets or intervene directly in credit markets. Brunnermeier also raised the possibility of separate news conferences for humans and machines, while warning that unfettered AI could weaken institutional trust and disadvantage people without advanced tools. The remarks came at the Kansas City Fed's annual economic symposium, where near-term issues including Treasury Secretary Scott Bessent's government debt buybacks and Fed Chairman Kevin Warsh's more direct policy communication received greater attention. Boston Fed President Susan Collins said AI and financial innovation create opportunities but also risks, including increased illicit activity. Other conference papers examined stablecoins and distributed-ledger technologies. Warsh presented AI more optimistically, focusing on whether it will produce a sustained productivity increase and affect workers, employment and monetary policy. Brunnermeier acknowledged potential benefits such as stronger risk management and oversight, but said addressing the dangers was necessary to secure those gains.