Thinking fast and slow: improving how we understand and control the use of LLMs

LLMs may sound like careful thinkers. But beneath the surface, there is just statistical pattern-matching. Financial services firms must stay sceptical, enforce human oversight, and design for safety. By Brandon Horwitz (with some help from ChatGPT)*, Chair of the IFoA’s AI Ethics, Governance and Risk Management Working Party. This is the fourth in a series of blogs by the working party.

Brandon Horwitz 19/11/25