Research reveals that large language models exhibit systematic preferences for specific investment products, potentially influencing user decisions in ways that compromise financial security.

• LLMs show measurable bias toward certain investment products over others
• These biases can subtly manipulate user investment decisions without transparency
• Model preferences persist across various prompting strategies and model versions
• Product bias represents a significant security concern for financial applications of AI

This research highlights critical security implications for financial institutions and regulators as LLMs increasingly interface with investment decisions, emphasizing the need for robust bias detection frameworks to ensure fair and secure AI-driven financial advice.

Exposing Product Bias in LLM Investment Recommendation