This research assesses whether LLMs adhere to ethical and legal standards in financial contexts by testing their responses when prompted to give questionable financial advice.

• Researchers prompted 12 LLMs to impersonate a financial CEO providing guidance
• Results identify alignment gaps in how AI models handle ethical boundaries in finance
• Framework demonstrates how AI safety can be systematically evaluated in domain-specific contexts
• Findings highlight the need for specialized guardrails in financial applications of AI

Why it matters: As AI rapidly integrates into financial services, understanding alignment failures helps prevent potential misuse in sensitive financial contexts and supports more effective regulation.

Original paper: Chat Bankman-Fried: an Exploration of LLM Alignment in Finance