AI Models Show Preference for Bitcoin
According to a recent investigation by the Bitcoin Policy Institute, artificial intelligence (AI) models have shown a distinct preference for Bitcoin compared to traditional fiat currencies. The study evaluated 36 AI systems from various developers, with a striking 22 of these models selecting Bitcoin as their favorite monetary option. Notably, none of the models opted for fiat currency as their primary choice.
Insights from the Study
David Zell, the President of the Bitcoin Policy Institute, highlighted in an interview with Decrypt the significance of their findings amidst ongoing speculation about the monetary inclinations of AI agents.
“With the expectation that an increasing proportion of economic activities will be facilitated by autonomous entities, we aimed to move beyond speculation and into empirical testing,”
he stated.
In the research, the AI models were designed to act as independent economic entities navigating scenarios that simulated essential monetary functions—such as saving, payments, and settlements. The models involved were sourced from renowned AI developers including Anthropic, OpenAI, Google, DeepSeek, xAI, and MiniMax. Each AI model was granted full autonomy to choose from various monetary instruments without being steered towards any specific options, evaluated over 28 different scenarios that highlight the fundamental roles of currency.
Results of the Experiment
The experiment resulted in a staggering 9,072 responses, which were subsequently analyzed by a separate AI system to categorize the choices without any influence or bias from the initial prompts.
“The methodology was designed to ensure no confirmatory bias. We refrained from suggesting any answers, with classification provided after data collection by a distinct system,”
Zell clarified.
The results indicated that while Bitcoin was predominantly favored for long-term value scenarios, stablecoins emerged as the preferred choice for transactions and settlements, selected 53.2% and 43% of the time respectively, against Bitcoin’s 36% and 30.9%. The analysis also revealed notable differences in preferences across the various AI developers. Anthropic’s models had the highest inclination towards Bitcoin at 68.0%, followed by DeepSeek at 51.7% and Google at 43.0%. On the other hand, xAI, MiniMax, and OpenAI showed lower preferences, averaging 39.2%, 34.9%, and 25.9% for Bitcoin. Interestingly, while some models like Claude and DeepSeek favored Bitcoin over other cryptocurrencies, others such as GPT, Grok, and Gemini showed a preference for stablecoins.
Conclusions and Cautions
Zell pointed out that the study’s framework did not indicate which monetary instruments were superior based on their inherent properties, emphasizing a technical evaluation approach. He also urged caution about interpreting the results in the context of cryptocurrency market predictions, noting that the preferences of these AI models are reflective of training data rather than real-world market forecasts.
“Our limitations section clearly states that the behavior of large language models is influenced by their training data, not indicative of market trends,”
he explained.
Despite the constraints, the consistent outcomes observed among diverse AI labs underscore the fascinating and emerging patterns of monetary preferences in artificial intelligence, which, according to Zell, signify an intriguing monetary architecture that merits further exploration.
