Michael Hsu, the US’s former acting comptroller of the currency, has called for banks and regulators to rethink how they handle the explainability challenge around large language models (LLMs). He wants to see them take a more decision-focused approach that exploits the unique nature of LLMs.

Explainability – the ability to follow the logic that leads to the outputs from artificial intelligence – has been one of the biggest hurdles to LLM adoption among banks. By design, LLMs involve a vast