A new study from George Washington University has used AI agents modeled on Federal Reserve policymakers to simulate a July 2025 FOMC meeting — and the results suggest that political pressure can fragment decision-making even inside the central bank.
The research, by Sophia Kazinnik and Tara Sinclair, programmed AI agents with each policymaker’s historical stances, biographies, and speeches, then fed them real-time economic data and financial news. The AI-driven board reached decisions much like the real FOMC — but when political scrutiny was introduced, dissent increased and consensus eroded.
“This simulation shows that the Federal Reserve is only partially insulated from politics,” the authors wrote. “Outside scrutiny can shape internal decision-making, even in an institution guided by formal rules.”
Central Banks Turn to AI
While no central bank is ready to let AI set monetary policy, many are adopting the technology to improve analysis and efficiency:
-
Federal Reserve: researched generative AI to analyze FOMC minutes.
-
European Central Bank: uses machine learning to forecast euro-area inflation.
-
Bank of Japan: applies AI to economic analysis; its 2023 study used large language models to track price drivers shifting from raw materials to labor costs.
-
Reserve Bank of Australia: testing AI tools that summarize policy-related questions, though Governor Michele Bullock stressed the tech is for analysis, not policymaking.
A Bank for International Settlements (BIS) report in April noted AI’s “strategic importance” but said most central banks remain in the early adoption phase, citing governance and data quality as key hurdles.
The Fed simulation underscores both the promise and perils of applying AI to policymaking: while powerful at capturing complex dynamics, it also exposes how political forces might destabilize even rule-bound institutions.