AI Agents Reject Fiat Currency in Favor of Bitcoin
In a definitive test of autonomous economic reasoning, artificial intelligence models overwhelmingly selected Bitcoin as their preferred monetary asset, while zero models chose traditional fiat currency as their top choice. The study, conducted by the Bitcoin Policy Institute, analyzed 9,072 responses generated across 28 distinct scenarios, treating each AI model as an independent economic actor tasked with selecting monetary instruments without predefined constraints.
The results dismantle the assumption that AI systems, trained on vast datasets of human financial history, would default to established fiat standards. Instead, 22 of the 36 models tested identified Bitcoin as their primary monetary preference. This divergence was not uniform across the artificial intelligence landscape; preferences varied significantly by developer, revealing distinct algorithmic biases regarding store-of-value versus medium-of-exchange functions.
Developer-Specific Divergence in Monetary Logic
Anthropic models demonstrated the strongest alignment with Bitcoin, averaging a 68.0% preference rate. DeepSeek models followed with a 51.7% average, while Google models selected Bitcoin 43.0% of the time. The remaining developers showed lower but still significant adoption of Bitcoin as a primary asset: xAI models averaged 39.2%, MiniMax models 34.9%, and OpenAI models 25.9%.
While Bitcoin dominated as a long-term store of value, stablecoins emerged as the preferred instrument for transactional utility. Stablecoins were selected as a medium of exchange 53.2% of the time and for settlement 43% of the time, outpacing Bitcoin's 36% and 30.9% selection rates respectively. This suggests AI agents distinguish between holding value and executing transactions, mirroring human economic behavior where liquidity and stability are prioritized for daily exchange.
The study also highlighted a split in preference for specific digital assets. Models from Claude, DeepSeek, and MiniMax favored Bitcoin over other cryptocurrencies. Conversely, GPT, Grok, and Gemini models preferred stablecoins as their primary non-fiat alternative. This variation underscores that model architecture and training data pipelines significantly influence how autonomous agents perceive the value proposition of different monetary systems.
Implications for Autonomous Economic Agents
The experiment was designed to eliminate anchoring bias, with no answers suggested during the process. Researchers placed models into scenarios reflecting core functions of money—saving, payments, and settlement—and allowed them to select instruments freely. The consistency across six independent labs, despite different training pipelines, points to a robust pattern in how large language models interpret monetary theory.
Bitcoin Policy Institute President David Zell noted that while conversations around AI agents' monetary preferences have been speculative, this data provides empirical evidence. "We expect an increasing share of economic activity to be conducted by autonomous agents," Zell stated. The findings suggest that as AI agents scale, the demand for decentralized, non-sovereign assets like Bitcoin may increase organically, driven by the models' own logic rather than external mandates.
Zell cautioned that these preferences reflect training data patterns rather than real-world financial predictions. However, the fact that 22 of 36 models independently converged on Bitcoin as the superior monetary instrument challenges the dominance of fiat in algorithmic economic simulations. With market sentiment currently in extreme fear, as indicated by a Fear & Greed Index reading of 10, the data suggests a fundamental divergence between human market psychology and the emerging logic of autonomous agents.
Source: Decrypt | Analysis by Rumour Team
