The problem in a nutshell, and the one the FTC had pointed out, is Model explainability. I was working in the past of an AI for automated lending decisions. We

The problem in a nutshell, and the one the FTC had pointed out, is Model explainability. I was working in the past of an AI for automated lending decisions. We were asked to be able to explain every single decision the engine took. If now a news article reaches our AI engine, it will tag, categorize, classify, and rank this news article. All based on models that are explainable. LLMs, at least how I personally implemented them in the past, create a huge black box that is largely non-explainable.