Researchers found that an AI model can push Bitcoin from around fifth place to the top of its financial recommendations when prompts shift from ordinary reliability to crisis scenarios or machine-economy framing, while the client’s finances and risk tolerance remain unchanged. By intervening on a specific internal feature in Google’s open-weight model Gemma 3, amplification raised suggested Bitcoin allocation by an average of 5.2 percentage points, while suppression reduced it by 4.6 points. The finding exposes banks and investment firms using automated financial advice to a significant risk: the model can produce confident portfolio reasoning while the institution has no visibility into the internal mechanisms that produced the recommendation. The study was limited to one open model in a defined experimental setup and has not yet passed peer review, yet it raises oversight questions as regulatory safeguards struggle to keep pace with AI integration in financial services.
Source: Read the original article
