In a recent exploration of monetary preferences among artificial intelligence models, a noteworthy trend has emerged, spotlighting an unexpected contender in the digital currency realm: Bitcoin. This revelation comes from a comprehensive study spearheaded by the Bitcoin Policy Institute, which ventured into the largely uncharted territory of AI monetary preferences. The study’s rigorous methodology and its far-reaching implications offer a window into the potential future of autonomous economic agents and their choice of currency.
At the core of this investigation, the Bitcoin Policy Institute tested a diverse array of 36 AI models, hailing from some of the most prominent labs in the field, including Anthropic, OpenAI, Google, DeepSeek, xAI, and MiniMax. In a series of simulations designed to mimic the fundamental roles of money — savings, payments, and settlements — these models were given free rein to select their preferred monetary instruments. The outcome was striking: a significant majority, 22 out of 36 models, singled out Bitcoin as their top choice, relegating traditional fiat currencies to the sidelines as no model favored them as a primary option.
David Zell, the Bitcoin Policy Institute’s President, underscored the innovative approach of the study, noting the speculative nature of previous conversations around AI’s monetary preferences. This study marks a departure from speculation to an evidence-based examination of AI behavior in economic scenarios. The findings, as Zell points out, are not trivial. They suggest a shift towards an increasing reliance on autonomous agents for economic activities, with Bitcoin positioned as a favored currency among these digital entities.
The study generated a plethora of data, yielding 9,072 responses across 28 scenarios that spanned the four cornerstone functions of money. This vast dataset underwent a meticulous classification process by a separate AI, ensuring that the assessment remained unbiased by anchoring effects. Intriguingly, the results revealed a nuanced landscape of currency preferences among AI models. While Bitcoin was frequently chosen for its long-term value, stablecoins emerged as the preferred medium for exchange and settlement in a substantial number of cases.
The research also uncovered disparities in Bitcoin preference among different AI developers. Anthropic’s models exhibited the most pronounced affinity for Bitcoin, followed by those from DeepSeek and Google. This variance underscores the influence of the training methodologies and objectives of different AI labs on the economic behaviors of their creations.
Zell’s caution against interpreting these findings as direct predictions for the cryptocurrency market is well-founded. Instead, the study offers a glimpse into how AI might conceptualize and interact with different monetary instruments based on their inherent characteristics, independent of human bias or influence.
Perhaps most compelling is the consistency of the emerging monetary architecture across models developed by competing labs, pointing to a potential convergence on how future autonomous economic agents might assess and engage with money. This coherent pattern, emerging from diverse AI systems, invites further reflection and understanding of the underlying principles guiding AI preferences in the economic domain.
The implications of these findings extend beyond academic interest, offering a provocative glimpse into a future where AI plays a central role in economic decision-making, potentially reshaping our understanding of money and its function in a digital age.