JACKSON HOLE, Wyoming, Aug 31 (Reuters) - Imagine a world where AI-powered trading is so prevalent, fast and effective that the Federal Reserve must give two news conferences, one for humans and another for machines, to avoid humans being completely shut out of responding to shifts in monetary policy.
Or where central bankers, who in recent decades have moved toward more transparency, must speak in opaque terms and act in unpredictable ways to keep AI agents from gaming the system.
Or where monetary policymakers must intervene directly in credit markets and build larger central bank balance sheets -- both ideas antithetical to the preferences of Fed Chairman Kevin Warsh and many of his counterparts -- to avoid the risk of AI-driven market manipulation.
Those were just a few of the dystopian scenarios that Princeton University economist Markus Brunnermeier laid out to global central bankers on Saturday in Jackson Hole, Wyoming, in what amounted to a plea for urgent action to prepare for and counter what he predicts could be sweeping disruption from the adoption of artificial intelligence in the financial world.
It was one of the final presentations at the Kansas City Fed's annual economic symposium, where far more immediate concerns like the market fallout from Treasury Secretary Scott Bessent's government debt buybacks and Warsh's notable shift to a franker style of policy communication dominated much of the conversation.
And it put a fresh market-focused lens on the potential implications of AI for central bankers, who have tended to focus on how the technology may change labor markets, enhance productivity, reduce inflation, or all three and more.
In Brunnermeier's telling, AI is so much better at processing information than humans that it will soon have the advantage of knowing with near certainty what central bankers will do even before those policymakers know it themselves, and can therefore devise trading and other strategies to circumvent those actions and turn a profit -- legally or otherwise.
When deployed at scale, AI agents could easily outmaneuver financial authorities and remake how financial markets behave, he argued, using the term "asymmetric understanding" to capture that dynamic.
"In a world in which humans interact with AI agents, a tail scenario may arise in which markets become less informative and more erratic," he wrote in a paper that was heavy on the philosophical and mostly devoid of the equations that typically pepper Jackson Hole presentations. "To prepare for such a scenario, regulation must shift toward simplicity and robustness. Central banks, in turn, face a game in which the market understands them better than they understand it."
Central bankers may feel they need to respond by speaking with less clarity, undoing a decades-long evolution toward straightforward communication of their views that has been widely credited with easing volatility and contributing to more efficient markets, he said.
"Transparency has to be rethought as predictability, arms the opponent in the financial dominance game and invites 'moves' that trap the public authorities," Brunnermeier wrote. "Hence, there is a case for more opacity."
Unfettered AI, he warned, could undermine trust in institutions and disadvantage those without the most sophisticated tools.
"Remedies such as two press conferences, one carrying an anchoring narrative for humans and one addressed to machines in the form of training data, or treating AI agents as influencers, will only partially alleviate this challenge," he said.
Symposium attendees speaking on the sidelines of the two-day conference said the paper helps frame the question of how AI will reshape financial markets and, in turn, their own behavior.
"AI and financial innovations provide real opportunities, and there are certainly some areas where there are risks where we would want to make sure that we are not seeing increases in illicit activities," Boston Fed President Susan Collins said in an interview. "I think we need to think through all of those things and papers like that one that help frame stark alternatives can facilitate a robust conversation."
Even so, Collins and other central bankers in Jackson Hole are for now less focused on a possibly fanciful dystopic future and more on the near-term and real-world implications of AI and financial innovation more generally.
Other papers at the conference took up related topics from a mundane and practical perspective -- how the emergence of stablecoins and distributed-ledger technologies could change central banking and the real economy, for instance, and how regulation should proceed to keep it safe.
Warsh in his Friday keynote notably had an optimistic take on AI as he sketched out what, for him, are the big questions.
"Will the application of AI cause a significant, sustained rise in productivity across the economy? And if so, when?" Warsh said. "What are the broad implications for workers and for the employment side of the Fed’s mandate?"
"We recognize that AI is a new variable—potentially a new factor of production—that will have consequences for both the economy and the conduct of monetary policy," Warsh said, without saying exactly how he expects that to play out.
Brunnermeier's paper did nod to AI's positive potential, including for better risk management and oversight sharper than any human could be.
But its focus was squarely on the potential dangers as, he wrote, "the best way to lock in the upside and avoid the downside."
Reporting by Ann Saphir with additional reporting by Howard Schneider; Editing by Chizu Nomiyama
