The downfall of many LLM approaches seems to be the "black box" nature of the output. Regulators are demanding more transparency in how these systems generate content. Could this be a credible answer?